CN116521761B - Sensor operation behavior mining method and system based on artificial intelligence - Google Patents
Sensor operation behavior mining method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN116521761B CN116521761B CN202310719547.6A CN202310719547A CN116521761B CN 116521761 B CN116521761 B CN 116521761B CN 202310719547 A CN202310719547 A CN 202310719547A CN 116521761 B CN116521761 B CN 116521761B
- Authority
- CN
- China
- Prior art keywords
- data
- feature vector
- data segment
- typical
- estimated
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000005065 mining Methods 0.000 title claims abstract description 121
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 38
- 239000013598 vector Substances 0.000 claims abstract description 450
- 230000002776 aggregation Effects 0.000 claims abstract description 39
- 238000004220 aggregation Methods 0.000 claims abstract description 39
- 230000006399 behavior Effects 0.000 claims description 263
- 238000004458 analytical method Methods 0.000 claims description 96
- 238000013507 mapping Methods 0.000 claims description 28
- 238000012512 characterization method Methods 0.000 claims description 26
- 238000005457 optimization Methods 0.000 claims description 13
- 230000009471 action Effects 0.000 claims description 11
- 230000004927 fusion Effects 0.000 claims description 10
- 239000012634 fragment Substances 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 238000012098 association analyses Methods 0.000 claims description 3
- 230000003542 behavioural effect Effects 0.000 claims description 2
- 238000012545 processing Methods 0.000 description 14
- 238000004364 calculation method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000007499 fusion processing Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2465—Query processing support for facilitating data mining operations in structured databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Mathematical Physics (AREA)
- Fuzzy Systems (AREA)
- Animal Behavior & Ethology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a sensor operation behavior mining method and system based on artificial intelligence, and relates to the technical field of artificial intelligence. In the invention, the data segment aggregation feature vector of the prior estimated data segment is determined; according to the data segment aggregate feature vector, determining the current focus middle feature vector; determining a data unit to be mined according to the current focusing intermediate feature vector; digging out the data feature vector to be mined corresponding to the data unit to be mined, and fusing to form a fused data feature vector; digging out an overall behavior feature vector corresponding to the operation behavior information to be analyzed, fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimating a later estimated data segment; the operation behavior profile data to be corresponding are formed based on the preceding estimation data piece and the following estimation data piece. Based on the above, the reliability of behavior mining can be improved.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a sensor operation behavior mining method and system based on artificial intelligence.
Background
Artificial intelligence (Artificial Intelligence, AI for short) is a theory, method, technique and application system that simulates, extends and extends human intelligence, senses environment, obtains knowledge and uses knowledge to obtain optimal results using digital computers or digital computer controlled computations. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning, automatic driving, intelligent traffic and other directions. Among them, machine learning and deep learning generally include artificial neural networks, confidence networks, reinforcement learning, transfer learning, induction learning, teaching learning, and the like.
In many applications of artificial intelligence, it is generally involved to mine some key information from a large amount of operation behavior data to perform characterization, that is, obtain some generalized content, so that a corresponding user can more quickly and effectively control operation behaviors. However, in the prior art, there is a problem that reliability is poor in the course of performing behavior mining.
Disclosure of Invention
In view of the above, the present invention aims to provide a method and a system for mining sensor operation behavior based on artificial intelligence, so as to improve reliability of behavior mining.
In order to achieve the above purpose, the embodiment of the present invention adopts the following technical scheme:
an artificial intelligence based sensor operation behavior mining method, comprising:
determining a data segment aggregation feature vector of a previous estimated data segment, which is already estimated by operational behavior information to be analyzed, wherein the operational behavior information to be analyzed is used for describing the operation of a sensor to be analyzed;
determining a current focusing intermediate feature vector for estimating a later estimated data segment corresponding to the earlier estimated data segment according to the data segment aggregation feature vector;
loading the current focusing intermediate feature vector to a first number of pending data unit analysis models, so as to determine to-be-mined data units in the to-be-analyzed operation behavior information according to the current focusing intermediate feature vector in the first number of pending data unit analysis models, wherein the first number is greater than 1;
mining data feature vectors to be mined corresponding to the data units to be mined, which are respectively determined based on the first number of data unit analysis models, in the operation behavior information to be analyzed, and fusing to form fused data feature vectors corresponding to the first number of data feature vectors to be mined;
Digging out an overall behavior feature vector corresponding to the running behavior information to be analyzed, fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimating the later estimated data segment corresponding to the focusing behavior feature vector;
and based on the previous estimated data segment and the subsequent estimated data segment, combining to form operation behavior profile data corresponding to the operation behavior information to be analyzed, wherein the operation behavior profile data is used for carrying out generalized description on the operation behavior information to be analyzed.
In some preferred embodiments, in the artificial intelligence based sensor operation behavior mining method, the step of loading the current focusing intermediate feature vector into a first number of pending data unit analysis models to determine the data units to be mined in the operation behavior information to be analyzed according to the current focusing intermediate feature vector, respectively, includes:
loading the current focusing intermediate feature vector to load the current focusing intermediate feature vector into a first number of pending data unit analysis models;
Determining a second number of data units to be confirmed in the operation behavior information to be analyzed according to the current focusing intermediate feature vector in an a-th data unit analysis model in the first number of data unit analysis models, and mining out a data feature vector to be confirmed corresponding to each data unit to be confirmed, wherein a is smaller than or equal to the first number, the second number of data units to be confirmed is a data unit to be confirmed except for data units to be mined, which are determined by the b-th data unit analysis model, in a third number of data units to be confirmed, the sum of b and 1 is equal to a, and the difference between the third number and a is equal to the difference between the second number and 1;
vector mapping operation is carried out on the second number of data feature vectors to be confirmed so as to output the possibility characterization parameters corresponding to the second number of data units to be confirmed respectively;
and analyzing the data units to be mined in the second number of data units to be confirmed according to the possibility characterization parameters corresponding to the second number of data units to be confirmed, wherein the data units to be mined corresponding to the first number of data unit analysis models are different from each other.
In some preferred embodiments, in the artificial intelligence based sensor operation behavior mining method, the step of aggregating feature vectors according to the data segments to determine a current focus intermediate feature vector for estimating a post-estimated data segment corresponding to the pre-estimated data segment includes:
determining a preceding focus intermediate feature vector for estimating the preceding estimated data segment;
and carrying out associated focusing characteristic analysis operation on the prior focusing intermediate characteristic vector and the data segment aggregation characteristic vector by utilizing a focusing characteristic analysis model so as to form a current focusing intermediate characteristic vector for estimating the post estimated data segment corresponding to the prior estimated data segment.
In some preferred embodiments, in the above method for mining sensor operation behaviors based on artificial intelligence, the step of mining, in the operation behavior information to be analyzed, feature vectors of data to be mined corresponding to the data units to be mined determined based on the first number of data unit analysis models, and fusing the feature vectors of fused data corresponding to the feature vectors of the data to be mined to form the first number of data units to be mined includes:
Mining data feature vectors to be mined corresponding to the data units to be mined, which are respectively determined based on the first number of data unit analysis models, in the operation behavior information to be analyzed;
vector association analysis operation is carried out on the first number of data feature vectors to be mined and the current focusing intermediate feature vector respectively so as to form a corresponding first number of association characterization parameters;
performing parameter mapping operation on the first number of relevance characterization parameters to form mapping relevance characterization parameters corresponding to the first number of data feature vectors to be mined;
and according to the mapping relevance characterization parameters corresponding to the first number of data feature vectors to be mined, performing superposition operation of corresponding weights on the first number of data feature vectors to be mined to form corresponding fusion data feature vectors.
In some preferred embodiments, in the artificial intelligence-based sensor operation behavior mining method, the step of mining the overall behavior feature vector corresponding to the operation behavior information to be analyzed, and fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimating the post-estimated data segment corresponding to the focusing behavior feature vector includes:
Digging out an overall behavior feature vector corresponding to the operation behavior information to be analyzed, and fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector;
determining a previous intermediate feature vector corresponding to the previous estimated data segment through a data segment estimation model, and performing vector aggregation analysis operation on the previous intermediate feature vector and the focusing behavior feature vector to estimate a fourth number of to-be-estimated data segments and data segment probability parameters corresponding to each to-be-estimated data segment;
and marking one undetermined estimated data segment with the maximum corresponding data segment probability parameter in the fourth number of undetermined estimated data segments to be marked as a later estimated data segment.
In some preferred embodiments, in the artificial intelligence based sensor operation behavior mining method, the step of determining the data segment aggregate feature vector of the previously estimated data segment for which the operation behavior information to be analyzed has been estimated includes:
digging out the integral feature vector of the data unit corresponding to the operation behavior information to be analyzed;
Performing key information mining operation on the prior estimated data segment of which the operation behavior information to be analyzed is estimated to form a data segment mining feature vector corresponding to the prior estimated data segment;
and determining the data segment aggregation feature vector of the prior estimated data segment based on the data unit integral feature vector and the data segment mining feature vector.
In some preferred embodiments, in the artificial intelligence based sensor operation behavior mining method, the step of mining the overall feature vector of the data unit corresponding to the operation behavior information to be analyzed includes:
determining a third number of data units to be confirmed corresponding to the operation behavior information to be analyzed; and mining a data unit mining feature vector corresponding to each data unit to be confirmed in the operation behavior information to be analyzed; vector aggregation operation is carried out on the third number of data unit mining feature vectors corresponding to the third number of data units to be confirmed so as to form the data unit integral feature vectors corresponding to the operation behavior information to be analyzed;
the step of determining the data segment aggregate feature vector of the prior estimated data segment based on the data unit overall feature vector and the data segment mining feature vector comprises the following steps:
Performing cascading combination operation on the data unit integral feature vector and the data segment mining feature vector to form a data segment aggregation feature vector of the previous estimated data segment; or determining a previous intermediate feature vector of the previous estimated data segment, and performing cascade combination operation on the previous intermediate feature vector, the data unit whole feature vector and the data segment mining feature vector to form a data segment aggregation feature vector of the previous estimated data segment.
In some preferred embodiments, in the above artificial intelligence based sensor operation behavior mining method, the sensor operation behavior mining method further includes:
extracting typical operation behavior information, and determining typical operation behavior profile data corresponding to the typical operation behavior information, wherein the typical operation behavior profile data comprises typical preceding data fragments and typical following data fragments;
estimating a typical previous estimated data segment of the typical operation behavior information through a candidate data segment estimation network, and mining a typical data segment aggregation feature vector corresponding to the typical previous estimated data segment;
Determining a typical focus intermediate feature vector for estimating a typical post-estimated data segment corresponding to the typical pre-estimated data segment according to the typical data segment aggregate feature vector through a candidate focus feature analysis model included in the candidate data segment estimation network, and loading the typical focus intermediate feature vector into a first number of candidate pending data unit analysis models to determine typical data units to be mined in the typical operation behavior information according to the typical focus intermediate feature vector through the first number of candidate pending data unit analysis models;
in the typical operation behavior information, mining typical data feature vectors corresponding to the typical data units to be mined, which are respectively determined based on the first number of candidate data unit analysis models, and fusing to form typical fusion data feature vectors corresponding to the first number of typical data feature vectors to be mined;
mining out a typical overall behavior feature vector corresponding to the typical operation behavior information, fusing the typical overall behavior feature vector in the typical fusion data feature vector to form a typical focusing behavior feature vector, and estimating a first output of a typical previous estimated data segment corresponding to the typical focusing behavior feature vector;
Digging out a typical intermediate feature vector of the typical focusing action feature vector, and estimating a second output of a typical previous estimated data segment corresponding to the typical operation action information based on the typical focusing intermediate feature vector and the typical intermediate feature vector;
and performing network optimization operation on the candidate data segment estimation network based on the first output of the typical previous estimation data segment, the second output of the typical previous estimation data segment and the typical subsequent data segment to form a corresponding optimized data segment estimation network.
In some preferred embodiments, in the artificial intelligence based sensor operation behavior mining method, the step of performing a network optimization operation on the candidate data segment estimation network based on the first output of the typical previous estimated data segment, the second output of the typical previous estimated data segment, and the typical subsequent data segment to form a corresponding optimized data segment estimation network includes:
determining a first index of learning cost based on the first output of the typical preceding estimated data segment and the distinguishing information between the typical following data segment;
Determining a second index of the learning cost based on the second output of the typical preceding estimated data segment and the distinguishing information between the typical following data segment;
determining a third index of learning cost based on distinguishing information between the first output of the typical prior estimated data segment and the second output of the typical prior estimated data segment;
and performing network optimization operation on the candidate data segment estimation network based on the first learning cost index and the second learning cost index, and performing network optimization operation on the candidate focusing characteristic analysis model in the candidate data segment estimation network based on the third learning cost index to form a corresponding optimized data segment estimation network.
The embodiment of the invention also provides an artificial intelligence-based sensor operation behavior mining system, which comprises a processor and a memory, wherein the memory is used for storing a computer program, and the processor is used for executing the computer program so as to realize the method.
The sensor operation behavior mining method and system based on artificial intelligence provided by the embodiment of the invention can determine the data segment aggregation feature vector of the prior estimated data segment; according to the data segment aggregate feature vector, determining the current focus middle feature vector; determining a data unit to be mined according to the current focusing intermediate feature vector; digging out the data feature vector to be mined corresponding to the data unit to be mined, and fusing to form a fused data feature vector; digging out an overall behavior feature vector corresponding to the operation behavior information to be analyzed, fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimating a later estimated data segment; the operation behavior profile data to be corresponding are formed based on the preceding estimation data piece and the following estimation data piece. Based on the foregoing, that is, the first number of pending data unit analysis models is adopted, a plurality of to-be-mined data units with related semantic content are determined from the to-be-analyzed operation behavior information, and then, through fusion processing, relatively more complete feature vectors can be formed, so that corresponding post-estimated data fragments can be estimated more reliably, and therefore, the reliability of behavior mining can be improved.
In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
FIG. 1 is a block diagram of an artificial intelligence based sensor operational behavior mining system according to an embodiment of the present invention.
Fig. 2 is a schematic flow chart of steps included in the artificial intelligence-based sensor operation behavior mining method according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of each module included in the artificial intelligence based sensor operation behavior mining apparatus according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in FIG. 1, an embodiment of the present invention provides an artificial intelligence based sensor operation behavior mining system. Wherein the artificial intelligence based sensor operation behavior mining system may include a memory and a processor.
In detail, the memory and the processor are electrically connected directly or indirectly to realize transmission or interaction of data. For example, electrical connection may be made to each other via one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that may exist in the form of software or firmware. The processor may be configured to execute the executable computer program stored in the memory, thereby implementing the artificial intelligence based sensor operation behavior mining method provided by the embodiment of the present invention.
It should be appreciated that in some possible embodiments, the Memory may be, but is not limited to, random access Memory (Random Access Memory, RAM), read Only Memory (ROM), programmable Read Only Memory (Programmable Read-Only Memory, PROM), erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), and the like.
It should be appreciated that in some possible embodiments, the processor may be a general purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a System on Chip (SoC), etc.; but also Digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
It should be appreciated that in some possible embodiments, the artificial intelligence based sensor operational behavioral mining system may be a server with data processing capabilities.
With reference to fig. 2, the embodiment of the invention further provides an artificial intelligence-based sensor operation behavior mining method, which can be applied to the artificial intelligence-based sensor operation behavior mining system. The method steps defined by the flow related to the artificial intelligence-based sensor operation behavior mining method can be realized by the artificial intelligence-based sensor operation behavior mining system.
The specific flow shown in fig. 2 will be described in detail.
Step S110, determining the data segment aggregation feature vector of the previous estimated data segment which is already estimated by the operation behavior information to be analyzed.
In the embodiment of the invention, the artificial intelligence-based sensor operation behavior mining system can determine the data segment aggregation feature vector of the previous estimated data segment which is estimated by the operation behavior information to be analyzed. The operation behavior information to be analyzed is used for describing the operation of the sensor to be analyzed, such as describing the operation behavior or operation process of the sensor to be analyzed, and the previously estimated data segment belongs to the data segment estimated in the previous flow.
Step S120, determining a current focusing middle feature vector for estimating a post-estimated data segment corresponding to the pre-estimated data segment according to the data segment aggregate feature vector.
In the embodiment of the invention, the artificial intelligence-based sensor operation behavior mining system can determine the current focusing intermediate feature vector for estimating the subsequent estimated data segment corresponding to the previous estimated data segment according to the data segment aggregation feature vector.
Step S130, performing loading processing on the current focusing intermediate feature vector, so as to load the current focusing intermediate feature vector into a first number of pending data unit analysis models, so as to determine to-be-mined data units in the to-be-analyzed operation behavior information according to the current focusing intermediate feature vector in the first number of pending data unit analysis models.
In the embodiment of the invention, the sensor operation behavior mining system based on artificial intelligence can load the current focusing middle feature vector to load the current focusing middle feature vector into a first number of pending data unit analysis models, so as to determine to-be-mined data units in the operation behavior information to be analyzed according to the current focusing middle feature vector in the first number of pending data unit analysis models. The first number may be greater than 1, i.e. a plurality of pending data unit analysis models.
And step S140, mining data feature vectors corresponding to the data units to be mined, which are determined based on the first number of data unit analysis models, in the operation behavior information to be analyzed, and fusing to form fused data feature vectors corresponding to the first number of data feature vectors to be mined.
In the embodiment of the invention, the sensor operation behavior mining system based on artificial intelligence can mine the data feature vectors to be mined corresponding to the data units to be mined, which are determined based on the first number of data unit analysis models, in the operation behavior information to be analyzed, and fuse the data feature vectors to be mined corresponding to the first number of data feature vectors to be mined, namely, fuse the first number of data feature vectors to be mined to form fused data feature vectors.
Step S150, the overall behavior feature vector corresponding to the operation behavior information to be analyzed is mined, the overall behavior feature vector is fused in the fused data feature vector to form a focusing behavior feature vector, and the post estimation data segment corresponding to the focusing behavior feature vector is estimated.
In the embodiment of the invention, the artificial intelligence-based sensor operation behavior mining system can mine out the overall behavior feature vector corresponding to the operation behavior information to be analyzed, fuse the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimate the later estimated data segment corresponding to the focusing behavior feature vector.
Step S160, based on the previous estimated data segment and the subsequent estimated data segment, operation behavior profile data corresponding to the operation behavior information to be analyzed is formed by combining.
In the embodiment of the invention, the artificial intelligence based sensor operation behavior mining system may combine to form operation behavior profile data corresponding to the operation behavior information to be analyzed based on the previous estimation data segment and the subsequent estimation data segment, that is, the first estimation data segment, the second estimation data segment, the third estimation data segment, the fourth estimation data segment and the like which are sequentially determined are combined to form the operation behavior profile data, and in addition, the data segments may refer to a sentence or a word and the like, and the specific granularity is not limited. The operation behavior profile data is used for generally describing the operation behavior information to be analyzed. Therefore, the effective compression of the operation behavior information to be analyzed can be realized, and the generalized and representative data can be obtained.
Based on the foregoing, that is, the first number of pending data unit analysis models is adopted, a plurality of to-be-mined data units with related semantic content are determined from the to-be-analyzed operation behavior information, and then, through fusion processing, relatively more complete feature vectors can be formed, so that corresponding post-estimated data fragments can be estimated more reliably, and therefore, the reliability of behavior mining can be improved.
It should be appreciated that, in some possible embodiments, step S110 in the foregoing description, that is, the step of determining the data segment aggregate feature vector of the previously estimated data segment for which the operational behavior information to be analyzed has been estimated, may further include the following specific implementation matters:
digging out the integral feature vector of the data unit corresponding to the operation behavior information to be analyzed;
performing key information mining operation on the prior estimated data segment of which the operation behavior information to be analyzed is estimated to form a data segment mining feature vector corresponding to the prior estimated data segment, for example, performing feature space mapping operation on the prior estimated data segment through a feature space mapping network to form a corresponding data segment mining feature vector;
and determining the data segment aggregation feature vector of the prior estimated data segment based on the data unit integral feature vector and the data segment mining feature vector.
It should be understood that, in some possible embodiments, the step of mining the overall feature vector of the data unit corresponding to the operational behavior information to be analyzed may further include the following specific implementation contents:
Determining a third number of to-be-confirmed data units corresponding to the to-be-analyzed operation behavior information, for example, a plurality of to-be-confirmed data units, namely a third number of to-be-confirmed data units, can be determined in the to-be-analyzed operation behavior information, each to-be-confirmed data unit belongs to a part of the to-be-analyzed operation behavior information, and each two to-be-confirmed data units are different or not identical;
in the operation behavior information to be analyzed, mining a data unit mining feature vector corresponding to each data unit to be confirmed, for example, mapping operation of feature space can be performed on the data unit to be confirmed to form a corresponding data unit mining feature vector;
and performing vector aggregation operation on a third number of data unit mining feature vectors corresponding to the third number of data units to be confirmed to form a data unit overall feature vector corresponding to the operation behavior information to be analyzed, for example, performing cascade combination operation on the third number of data unit mining feature vectors to form a data unit overall feature vector, such as { data unit mining feature vector 1, data unit mining feature vector 2, data unit mining feature vector 3..+ -, or performing weighted superposition operation on the third number of data unit mining feature vectors based on weighting coefficients respectively configured by the third number of data units to be confirmed to obtain the data unit overall feature vector.
It should be appreciated that, in some possible embodiments, the step of determining the data segment aggregate feature vector of the previously estimated data segment based on the data unit global feature vector and the data segment mining feature vector may further include the following specific implementation matters:
performing cascading combination operation on the data unit integral feature vector and the data segment mining feature vector to form a data segment aggregation feature vector of the previous estimated data segment; or alternatively, the first and second heat exchangers may be,
determining a previous intermediate feature vector of the previous estimated data segment, wherein the previous estimated data segment can be obtained by performing estimation analysis based on the previous intermediate feature vector, and in the case that the subsequent estimated data segment belongs to the first data segment for estimation, one configured reference feature vector can be used as the previous intermediate feature vector of the previous estimated data segment;
and performing cascading combination operation on the previous intermediate feature vector, the data unit integral feature vector and the data segment mining feature vector to form a data segment aggregation feature vector of the previous estimated data segment, such as { the previous intermediate feature vector, the data unit integral feature vector, the data segment mining feature vector }.
It should be appreciated that, in some possible embodiments, the step S120 of determining the current focus intermediate feature vector for estimating the post-estimated data segment corresponding to the pre-estimated data segment according to the data segment aggregate feature vector may further include the following specific implementation matters described below:
determining a preceding focusing intermediate feature vector for estimating the preceding estimated data segment, wherein the preceding focusing intermediate feature vector of the preceding estimated data segment belongs to the preceding estimated data segment, and analyzing the preceding estimated data segment as a following estimated data segment to obtain a current focusing intermediate feature vector;
performing a correlation focus feature analysis operation on the previous focus intermediate feature vector and the data segment aggregate feature vector by using a focus feature analysis model to form a current focus intermediate feature vector for estimating a subsequent estimated data segment corresponding to the previous estimated data segment, wherein the correlation focus feature analysis operation may include performing a focus feature analysis operation on the data segment aggregate feature vector based on the previous focus intermediate feature vector to obtain a first focus result, performing a focus feature analysis operation on the previous focus intermediate feature vector based on the data segment aggregate feature vector to obtain a second focus result, and then, stacking or cascading the first focus result and the second focus result to form a current focus intermediate feature vector of the subsequent estimated data segment; in addition, the "performing the focus feature analysis operation on the data segment aggregate feature vector based on the preceding focus intermediate feature vector" may mean performing a mapping operation on the preceding focus intermediate feature vector based on a first mapping matrix and a second mapping matrix, and performing a mapping operation on the data segment aggregate feature vector based on a third mapping matrix, and then calculating a product between a transpose vector of the obtained first mapping vector and the obtained third mapping vector, and then weighting the obtained second mapping vector based on the product, so as to obtain a first focus result.
It should be understood that, in some possible embodiments, the step S130 of loading the current focus intermediate feature vector to be loaded into the first number of pending data unit analysis models to determine the data unit to be mined in the operation behavior information to be analyzed according to the current focus intermediate feature vector in the first number of pending data unit analysis models, may further include the following specific implementation matters described below:
loading the current focusing intermediate feature vector to load the current focusing intermediate feature vector into a first number of pending data unit analysis models, namely loading the current focusing intermediate feature vector into a plurality of pending data unit analysis models;
determining a second number of data units to be confirmed (at least part of the determined third number of data units to be confirmed) in the operation behavior information to be analyzed according to the current focusing middle feature vector in an a-th data unit analysis model (such as a first data unit analysis model, a second data unit analysis model, a third data unit analysis model and the like) in the first number of data unit analysis models, and mining a corresponding data feature vector of each data unit to be confirmed, wherein a is smaller than or equal to the first number, the second number of data units to be confirmed are data units to be confirmed other than the data units to be mined determined by the b-th data unit analysis model in the third number of data units to be confirmed, the sum of b and 1 is equal to a, and the difference between the third number and a is equal to the difference between the second number and 1;
Vector mapping operation is performed on the second number of feature vectors of data to be confirmed to output probability characterization parameters corresponding to each of the second number of data units to be confirmed, for example, the vector mapping operation may be implemented by a multi-layer Perceptron (MLP) unit, that is, the second number of feature vectors of data to be confirmed are loaded into the multi-layer Perceptron unit to analyze the probability characterization parameters corresponding to each data unit to be confirmed, or further processing may be performed on output data of the multi-layer Perceptron unit to obtain the probability characterization parameters, for example, processing is performed through an activation function, so that the output belongs to (0, 1);
according to the respective corresponding likelihood characterization parameters of the second number of to-be-confirmed data units, analyzing to-be-mined data units in the second number of to-be-confirmed data units, wherein the to-be-mined data units corresponding to the respective first number of to-be-confirmed data unit analysis models are different from each other, for example, the to-be-confirmed data unit with the maximum corresponding likelihood characterization parameter can be used as the to-be-mined data unit corresponding to the a-th to-be-confirmed data unit analysis model.
It should be understood that, in some possible embodiments, the data unit to be confirmed is obtained from the third number of data units to be confirmed through the a-th data unit analysis model, and the data unit to be mined in the operation behavior information to be analyzed is determined from the obtained data units to be confirmed. When the a-th pending data unit analysis model determines the data unit to be mined, deleting the data unit to be mined determined by the a-th pending data unit analysis model in the third number of data units to be confirmed, so that other pending data unit analysis models cannot determine the data unit to be mined determined by the a-th pending data unit analysis model again, obtaining a first number of data units to be mined with similar but incomplete semantic content, and improving the reliability of determining the data unit to be mined.
It should be understood that, in some possible embodiments, the step S140 of mining the feature vectors of the data units to be mined corresponding to the data units to be mined determined based on the first number of pending data unit analysis models in the operation behavior information to be analyzed, and fusing the feature vectors of the fused data corresponding to the first number of feature vectors of the data units to be mined may further include the following specific implementation matters:
Mining the data feature vectors to be mined corresponding to the data units to be mined, which are determined based on the first number of data unit analysis models, in the operation behavior information to be analyzed, namely mining key information of the data units to be mined to obtain the data feature vectors to be mined;
vector association analysis operation is carried out on the first number of data feature vectors to be mined and the current focusing intermediate feature vector respectively so as to form a corresponding first number of association characterization parameters;
performing parameter mapping operation on the first number of relevance characterization parameters to form mapping relevance characterization parameters corresponding to the first number of data feature vectors to be mined, wherein the parameters can be mapped to intervals (0, 1) through the parameter mapping operation;
and according to the mapping relevance characterization parameters corresponding to the first number of data feature vectors to be mined, performing superposition operation of corresponding weights on the first number of data feature vectors to be mined to form corresponding fusion data feature vectors, namely, taking the mapping relevance characterization parameters as weighting coefficients of the corresponding data feature vectors to be mined.
It should be understood that, in some possible embodiments, the step of performing vector correlation analysis on the first number of feature vectors of the data to be mined and the current focus intermediate feature vector to form a corresponding first number of correlation characterization parameters may include:
processing the first number of data feature vectors to be mined and the current focusing intermediate feature vector to form a vector with a consistent channel (dimension), and respectively performing vector superposition operation on the processed first number of data feature vectors to be mined and the processed current focusing intermediate feature vector, so that a first number of relevance characterization parameters can be formed; or,
acquiring a first configuration weighting coefficient corresponding to the data unit to be mined, and acquiring a second configuration weighting coefficient corresponding to the current focusing intermediate feature vector;
and respectively carrying out weighted calculation on the first number of data feature vectors to be mined according to the first configuration weighting coefficient to form a first number of weighted data feature vectors to be mined, carrying out weighted calculation on the current focusing middle feature vector according to the second configuration weighting coefficient to form a corresponding target weighting parameter, and then adding the target weighting parameter into each weighted data feature vector to be mined to form a first number of relevance characterization parameters.
It should be understood that, in some possible embodiments, the step S150 of mining the overall behavior feature vector corresponding to the running behavior information to be analyzed, and fusing the overall behavior feature vector in the fused data feature vector to form a focused behavior feature vector, and estimating the post-estimated data segment corresponding to the focused behavior feature vector may further include the following specific implementation matters described below:
the overall behavior feature vector corresponding to the operation behavior information to be analyzed is mined, for example, the overall operation behavior information to be analyzed can be subjected to feature space mapping operation to form a corresponding overall behavior feature vector, and the overall behavior feature vector is fused in the fused data feature vector to form a focusing behavior feature vector, for example, superposition operation is performed on the fused data feature vector and the overall behavior feature vector;
determining a previous intermediate feature vector corresponding to the previous estimated data segment through a data segment estimation model, and performing vector aggregation analysis operation on the previous intermediate feature vector and the focusing action feature vector to estimate a fourth number of pending estimated data segments and data segment possibility parameters corresponding to each pending estimated data segment, for example, the previous intermediate feature vector and the focusing action feature vector may be subjected to vector aggregation operation to form a corresponding pending aggregation feature vector, and then, analysis estimation may be performed based on the pending aggregation feature vector to determine a fourth number of pending estimated data segments (in suspected data segments) and data segment possibility parameters corresponding to each pending estimated data segment;
And marking one undetermined estimated data segment with the maximum corresponding data segment probability parameter in the fourth number of undetermined estimated data segments to be marked as a later estimated data segment.
It should be understood that, in some possible embodiments, the step of mining the overall behavior feature vector corresponding to the running behavior information to be analyzed may further include the following specific implementation contents:
performing word segmentation processing on the operation behavior information to be analyzed to form a behavior information word set corresponding to the operation behavior information to be analyzed, performing coding processing on each behavior information word in the behavior information word set to form a word vector corresponding to each behavior information word, and performing mapping processing on the behavior information word set based on the vector similarity between word vectors corresponding to every two behavior information words to form a corresponding target map, wherein in the target map, each map object corresponds to one behavior information word, the weight of a map line segment connecting two map objects corresponds to the vector similarity between word vectors corresponding to two behavior information words corresponding to the two map objects, and the relationship can be positive correlation;
Respectively or sequentially taking a map object corresponding to each behavior information word in the behavior information word set as a starting map object, determining an adjacent map object cluster corresponding to the starting map object in the target map, and starting from the starting map object, performing traversing operation on the adjacent map object cluster based on the weight of a map line segment to determine a first traversing map object, determining an adjacent map object cluster of the first traversing map object, and performing traversing operation such as the starting map object until traversing is stopped to form a traversing path, wherein the possibility of stopping in the traversing process is a preconfigured reference possibility;
based on the steps, after forming a plurality of traversal paths corresponding to each map object, determining one traversal path with the maximum weight mean value in the plurality of traversal paths corresponding to the map object as a corresponding target traversal path, and performing focusing feature analysis operation on the word vector corresponding to the map object based on the word vector corresponding to each behavior information word corresponding to the target traversal path respectively to obtain each focusing feature analysis feature vector corresponding to the map object, and performing superposition operation on each focusing feature analysis feature vector corresponding to the map object and the word vector corresponding to the behavior information word corresponding to the map object to form a superposition feature vector corresponding to the map object;
And cascading and combining the superposition feature vectors corresponding to each map object to form an overall behavior feature vector corresponding to the operation behavior information to be analyzed.
It should be understood that, in some possible embodiments, the step S150 of mining the overall behavior feature vector corresponding to the running behavior information to be analyzed, and fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimating the post-estimated data segment corresponding to the focusing behavior feature vector may further include the following specific implementation contents:
digging out the integral behavior feature vector corresponding to the operation behavior information to be analyzed, as described in the previous related description, and fusing the integral behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, as described in the previous related description;
performing a vector analysis operation on the focusing action feature vector and the current focusing intermediate feature vector by using a data segment estimation model to form primary estimation data and primary intermediate feature vectors, wherein the primary estimation data is obtained by performing an analysis estimation operation on the basis of the primary intermediate feature vector, for example, the focusing action feature vector and the current focusing intermediate feature vector can be subjected to an aggregation operation, such as superposition, cascade combination or associated focusing feature analysis operation, so as to form corresponding primary intermediate feature vectors, and then estimation analysis can be performed on the basis of the primary intermediate feature vectors so as to form corresponding primary estimation data, wherein the primary estimation data can comprise a fourth number of pending estimation data segments and a first data segment possibility parameter corresponding to each of the fourth number of pending estimation data segments;
Vector aggregation operation is carried out on the current focusing intermediate feature vector and the primary intermediate feature vector so as to form a corresponding aggregation intermediate feature vector;
performing analysis and estimation operation on the aggregate intermediate feature vector to output intermediate estimation data, wherein the intermediate estimation data can comprise a fourth number of data segments to be estimated and a second data segment possibility parameter corresponding to each data segment to be estimated in the fourth number of data segments to be estimated;
and performing data merging analysis operation on the primary estimation data and the intermediate estimation data to output the later estimation data segment corresponding to the focusing action feature vector, for example, the primary estimation data and the intermediate estimation data may be subjected to mean value or weighted mean value calculation to form corresponding target estimation data, the target estimation data may include a target data segment likelihood parameter corresponding to each of a fourth number of pending estimation data segments and the fourth number of pending estimation data segments, and then, the pending estimation data segment with the largest corresponding target data segment likelihood parameter may be used as the corresponding later estimation data segment.
Wherein, it should be understood that, in some possible embodiments, the step of performing a vector aggregation operation on the current focusing intermediate feature vector and the primary intermediate feature vector to form a corresponding aggregated intermediate feature vector may further include the following specific implementation matters:
performing linear processing on the current focusing intermediate feature vector by a vector linear processing unit (MLP as described above) to form a current linear intermediate feature vector of a preset channel;
the primary intermediate feature vectors are linearly integrated by a vector linear integration unit (MLP as described above) to form primary linear intermediate feature vectors of the preset channel;
and performing addition operation on the current linear intermediate feature vector and the primary linear intermediate feature vector to form a corresponding aggregate intermediate feature vector.
It should be appreciated that in some possible embodiments, the artificial intelligence based sensor operation behavior mining method may further include the following specific implementation:
extracting typical operation behavior information, and determining typical operation behavior profile data corresponding to the typical operation behavior information, wherein the typical operation behavior profile data comprises typical preceding data segments and typical following data segments (such as a first typical data segment, a second typical data segment, a third typical data segment and the like, wherein the first typical data segment is taken as the typical preceding data segment, the second typical data segment is taken as the typical following data segment, and the second typical data segment is taken as the typical preceding data segment, and the third typical data segment is taken as the typical following data segment;
Estimating a typical previous estimated data segment of the typical operation behavior information through a candidate data segment estimation network, and mining a typical data segment aggregation feature vector corresponding to the typical previous estimated data segment, as described in the previous correlation;
determining a typical focus intermediate feature vector for estimating a typical post-estimated data segment corresponding to the typical pre-estimated data segment according to the typical data segment aggregate feature vector through a candidate focus feature analysis model included in the candidate data segment estimation network, and loading the typical focus intermediate feature vector into a first number of candidate pending data unit analysis models to determine typical data units to be mined in the typical operation behavior information according to the typical focus intermediate feature vector through the first number of candidate pending data unit analysis models, respectively, as described in the foregoing;
in the typical operation behavior information, mining typical data feature vectors corresponding to the typical data units to be mined, which are respectively determined based on the first number of candidate data unit analysis models, and fusing to form typical fusion data feature vectors corresponding to the first number of typical data feature vectors to be mined, as described in the previous related description;
Digging out a typical overall behavior feature vector corresponding to the typical operation behavior information, fusing the typical overall behavior feature vector in the typical fusion data feature vector to form a typical focusing behavior feature vector, estimating a first output of a typical previous estimated data segment corresponding to the typical focusing behavior feature vector, namely performing estimation analysis operation based on the typical focusing behavior feature vector to obtain corresponding first output data, namely obtaining estimated data corresponding to a typical subsequent data segment, and referring to the related description of the primary estimated data in the previous description;
digging out typical intermediate feature vectors of the typical focusing action feature vectors, estimating second output of typical previous estimated data segments corresponding to the typical running action information based on the typical focusing intermediate feature vectors and the typical intermediate feature vectors, namely performing estimation analysis operation based on the typical focusing intermediate feature vectors and the typical intermediate feature vectors to obtain corresponding second output data, wherein the related description of the intermediate estimated data can be referred to in the foregoing;
based on the first output of the typical prior estimated data segment, the second output of the typical prior estimated data segment, and the typical subsequent data segment, performing a network optimization operation on the candidate data segment estimation networks to form corresponding optimized data segment estimation networks, which may be used to perform the aforementioned steps S110-S160.
It should be appreciated that in some possible embodiments, the step of performing a network optimization operation on the candidate data segment estimation network to form a corresponding optimized data segment estimation network based on the first output of the typical previous estimated data segment, the second output of the typical previous estimated data segment, and the typical subsequent data segment may further include the following detailed implementation:
determining a first index of learning cost, namely the difference between first estimated data and actual data, based on the first output of the typical previous estimated data segment and the distinguishing information between the typical subsequent data segment, and performing error calculation, wherein a specific function is not limited;
determining a second index of the learning cost, namely the difference between the second estimated data and the actual data, based on the second output of the typical previous estimated data segment and the distinguishing information between the typical subsequent data segment, and performing error calculation, wherein a specific function is not limited;
determining a third index of the learning cost, namely the difference between the first type of estimated data and the second type of estimated data, based on the distinguishing information between the first output of the typical previous estimated data segment and the second output of the typical previous estimated data segment, and performing error calculation, wherein a specific function is not limited;
And carrying out network optimization operation on the candidate data segment estimation network based on the first learning cost index and the second learning cost index, and carrying out network optimization operation on the candidate focusing characteristic analysis model in the candidate data segment estimation network based on the third learning cost index, namely carrying out optimization adjustment on network parameters so as to form a corresponding optimized data segment estimation network.
With reference to fig. 3, the embodiment of the invention further provides an artificial intelligence-based sensor operation behavior mining device, which can be applied to the artificial intelligence-based sensor operation behavior mining system. Wherein, the artificial intelligence based sensor operation behavior mining device may include:
the first feature vector determining module is used for determining a data segment aggregation feature vector of a previous estimated data segment which is estimated by the operation behavior information to be analyzed, wherein the operation behavior information to be analyzed is used for describing the operation of the sensor to be analyzed;
the second feature vector determining module is used for determining a current focusing middle feature vector for estimating the post-estimated data segment corresponding to the prior estimated data segment according to the data segment aggregation feature vector;
The data unit determining module is used for carrying out loading processing on the current focusing intermediate feature vector so as to load the current focusing intermediate feature vector into a first number of pending data unit analysis models, and determining data units to be mined in the running behavior information to be analyzed in the first number of pending data unit analysis models according to the current focusing intermediate feature vector respectively, wherein the first number is more than 1;
the feature vector fusion module is used for mining the feature vectors of the data to be mined corresponding to the data units to be mined, which are respectively determined based on the first number of the data unit analysis models, in the operation behavior information to be analyzed, and fusing the feature vectors of the data to be mined to form fused feature vectors of the data to be mined corresponding to the first number of the feature vectors of the data to be mined;
the data segment estimation module is used for mining out an integral behavior feature vector corresponding to the operation behavior information to be analyzed, fusing the integral behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimating the later estimated data segment corresponding to the focusing behavior feature vector;
and the behavior profile data determining module is used for combining and forming the operation behavior profile data corresponding to the operation behavior information to be analyzed based on the previous estimated data segment and the subsequent estimated data segment, and the operation behavior profile data is used for carrying out generalized description on the operation behavior information to be analyzed.
In summary, the method and the system for mining the sensor operation behavior based on the artificial intelligence provided by the invention can determine the data segment aggregation feature vector of the prior estimated data segment; according to the data segment aggregate feature vector, determining the current focus middle feature vector; determining a data unit to be mined according to the current focusing intermediate feature vector; digging out the data feature vector to be mined corresponding to the data unit to be mined, and fusing to form a fused data feature vector; digging out an overall behavior feature vector corresponding to the operation behavior information to be analyzed, fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimating a later estimated data segment; the operation behavior profile data to be corresponding are formed based on the preceding estimation data piece and the following estimation data piece. Based on the foregoing, that is, the first number of pending data unit analysis models is adopted, a plurality of to-be-mined data units with related semantic content are determined from the to-be-analyzed operation behavior information, and then, through fusion processing, relatively more complete feature vectors can be formed, so that corresponding post-estimated data fragments can be estimated more reliably, and therefore, the reliability of behavior mining can be improved.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. An artificial intelligence based sensor operation behavior mining method, comprising:
determining a data segment aggregation feature vector of a previous estimated data segment, which is already estimated by operational behavior information to be analyzed, wherein the operational behavior information to be analyzed is used for describing the operation of a sensor to be analyzed;
determining a current focusing intermediate feature vector for estimating a later estimated data segment corresponding to the earlier estimated data segment according to the data segment aggregation feature vector;
loading the current focusing intermediate feature vector to a first number of pending data unit analysis models, so as to determine to-be-mined data units in the to-be-analyzed operation behavior information according to the current focusing intermediate feature vector in the first number of pending data unit analysis models, wherein the first number is greater than 1;
Mining data feature vectors to be mined corresponding to the data units to be mined, which are respectively determined based on the first number of data unit analysis models, in the operation behavior information to be analyzed, and fusing to form fused data feature vectors corresponding to the first number of data feature vectors to be mined;
digging out an overall behavior feature vector corresponding to the running behavior information to be analyzed, fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector, and estimating the later estimated data segment corresponding to the focusing behavior feature vector;
based on the previous estimated data segment and the subsequent estimated data segment, combining to form operation behavior profile data corresponding to the operation behavior information to be analyzed, wherein the operation behavior profile data is used for carrying out generalized description on the operation behavior information to be analyzed;
the step of loading the current focusing intermediate feature vector to be loaded into a first number of pending data unit analysis models to determine to-be-mined data units in the to-be-analyzed operation behavior information according to the current focusing intermediate feature vector in the first number of pending data unit analysis models, includes:
Loading the current focusing intermediate feature vector to load the current focusing intermediate feature vector into a first number of pending data unit analysis models;
determining a second number of data units to be confirmed in the operation behavior information to be analyzed according to the current focusing intermediate feature vector in an a-th data unit analysis model in the first number of data unit analysis models, and mining out a data feature vector to be confirmed corresponding to each data unit to be confirmed, wherein a is smaller than or equal to the first number, the second number of data units to be confirmed is a data unit to be confirmed except for data units to be mined, which are determined by the b-th data unit analysis model, in a third number of data units to be confirmed, the sum of b and 1 is equal to a, and the difference between the third number and a is equal to the difference between the second number and 1;
vector mapping operation is carried out on the second number of data feature vectors to be confirmed so as to output the possibility characterization parameters corresponding to the second number of data units to be confirmed respectively;
and analyzing the data units to be mined in the second number of data units to be confirmed according to the possibility characterization parameters corresponding to the second number of data units to be confirmed, wherein the data units to be mined corresponding to the first number of data unit analysis models are different from each other.
2. The artificial intelligence based sensor operation behavior mining method according to claim 1, wherein the step of determining a current focus intermediate feature vector for estimating a subsequent estimated data segment corresponding to the previous estimated data segment based on the data segment aggregate feature vector comprises:
determining a preceding focus intermediate feature vector for estimating the preceding estimated data segment;
and carrying out associated focusing characteristic analysis operation on the prior focusing intermediate characteristic vector and the data segment aggregation characteristic vector by utilizing a focusing characteristic analysis model so as to form a current focusing intermediate characteristic vector for estimating the post estimated data segment corresponding to the prior estimated data segment.
3. The method for mining operation behaviors of a sensor based on artificial intelligence according to claim 1, wherein the step of mining the feature vectors of the data to be mined corresponding to the data units to be mined determined based on the analysis models of the first number of data units to be determined respectively in the operation behavior information to be analyzed, and fusing the feature vectors of the fused data corresponding to the feature vectors of the data to be mined to form the feature vectors of the first number of data units to be mined comprises:
Mining data feature vectors to be mined corresponding to the data units to be mined, which are respectively determined based on the first number of data unit analysis models, in the operation behavior information to be analyzed;
vector association analysis operation is carried out on the first number of data feature vectors to be mined and the current focusing intermediate feature vector respectively so as to form a corresponding first number of association characterization parameters;
performing parameter mapping operation on the first number of relevance characterization parameters to form mapping relevance characterization parameters corresponding to the first number of data feature vectors to be mined;
and according to the mapping relevance characterization parameters corresponding to the first number of data feature vectors to be mined, performing superposition operation of corresponding weights on the first number of data feature vectors to be mined to form corresponding fusion data feature vectors.
4. The artificial intelligence based sensor operation behavior mining method according to claim 1, wherein the step of mining out the overall behavior feature vector corresponding to the operation behavior information to be analyzed, and fusing the overall behavior feature vector in the fused data feature vector to form a focused behavior feature vector, and estimating the post-estimated data segment corresponding to the focused behavior feature vector, comprises:
Digging out an overall behavior feature vector corresponding to the operation behavior information to be analyzed, and fusing the overall behavior feature vector in the fused data feature vector to form a focusing behavior feature vector;
determining a previous intermediate feature vector corresponding to the previous estimated data segment through a data segment estimation model, and performing vector aggregation analysis operation on the previous intermediate feature vector and the focusing behavior feature vector to estimate a fourth number of to-be-estimated data segments and data segment probability parameters corresponding to each to-be-estimated data segment;
and marking one undetermined estimated data segment with the maximum corresponding data segment probability parameter in the fourth number of undetermined estimated data segments to be marked as a later estimated data segment.
5. The artificial intelligence based sensor operation behavior mining method according to claim 1, wherein the step of determining a data segment aggregate feature vector of a previous estimated data segment for which operation behavior information to be analyzed has been estimated, comprises:
digging out the integral feature vector of the data unit corresponding to the operation behavior information to be analyzed;
Performing key information mining operation on the prior estimated data segment of which the operation behavior information to be analyzed is estimated to form a data segment mining feature vector corresponding to the prior estimated data segment;
and determining the data segment aggregation feature vector of the prior estimated data segment based on the data unit integral feature vector and the data segment mining feature vector.
6. The artificial intelligence based sensor operation behavior mining method according to claim 5, wherein the step of mining the data unit integral feature vector corresponding to the operation behavior information to be analyzed comprises:
determining a third number of data units to be confirmed corresponding to the operation behavior information to be analyzed; and mining a data unit mining feature vector corresponding to each data unit to be confirmed in the operation behavior information to be analyzed; vector aggregation operation is carried out on the third number of data unit mining feature vectors corresponding to the third number of data units to be confirmed so as to form the data unit integral feature vectors corresponding to the operation behavior information to be analyzed;
the step of determining the data segment aggregate feature vector of the prior estimated data segment based on the data unit overall feature vector and the data segment mining feature vector comprises the following steps:
Performing cascading combination operation on the data unit integral feature vector and the data segment mining feature vector to form a data segment aggregation feature vector of the previous estimated data segment; or determining a previous intermediate feature vector of the previous estimated data segment, and performing cascade combination operation on the previous intermediate feature vector, the data unit whole feature vector and the data segment mining feature vector to form a data segment aggregation feature vector of the previous estimated data segment.
7. The artificial intelligence based sensor operation behavior mining method according to any one of claims 1 to 6, wherein the sensor operation behavior mining method further comprises:
extracting typical operation behavior information, and determining typical operation behavior profile data corresponding to the typical operation behavior information, wherein the typical operation behavior profile data comprises typical preceding data fragments and typical following data fragments;
estimating a typical previous estimated data segment of the typical operation behavior information through a candidate data segment estimation network, and mining a typical data segment aggregation feature vector corresponding to the typical previous estimated data segment;
Determining a typical focus intermediate feature vector for estimating a typical post-estimated data segment corresponding to the typical pre-estimated data segment according to the typical data segment aggregate feature vector through a candidate focus feature analysis model included in the candidate data segment estimation network, and loading the typical focus intermediate feature vector into a first number of candidate pending data unit analysis models to determine typical data units to be mined in the typical operation behavior information according to the typical focus intermediate feature vector through the first number of candidate pending data unit analysis models;
in the typical operation behavior information, mining typical data feature vectors corresponding to the typical data units to be mined, which are respectively determined based on the first number of candidate data unit analysis models, and fusing to form typical fusion data feature vectors corresponding to the first number of typical data feature vectors to be mined;
mining out a typical overall behavior feature vector corresponding to the typical operation behavior information, fusing the typical overall behavior feature vector in the typical fusion data feature vector to form a typical focusing behavior feature vector, and estimating a first output of a typical previous estimated data segment corresponding to the typical focusing behavior feature vector;
Digging out a typical intermediate feature vector of the typical focusing action feature vector, and estimating a second output of a typical previous estimated data segment corresponding to the typical operation action information based on the typical focusing intermediate feature vector and the typical intermediate feature vector;
and performing network optimization operation on the candidate data segment estimation network based on the first output of the typical previous estimation data segment, the second output of the typical previous estimation data segment and the typical subsequent data segment to form a corresponding optimized data segment estimation network.
8. The artificial intelligence based sensor operational behavioral mining method of claim 7, wherein said step of performing a network optimization operation on said candidate data segment estimation network based on a first output of said typical prior estimated data segment, a second output of said typical prior estimated data segment, and said typical subsequent data segment to form a corresponding optimized data segment estimation network comprises:
determining a first index of learning cost based on the first output of the typical preceding estimated data segment and the distinguishing information between the typical following data segment;
Determining a second index of the learning cost based on the second output of the typical preceding estimated data segment and the distinguishing information between the typical following data segment;
determining a third index of learning cost based on distinguishing information between the first output of the typical prior estimated data segment and the second output of the typical prior estimated data segment;
and performing network optimization operation on the candidate data segment estimation network based on the first learning cost index and the second learning cost index, and performing network optimization operation on the candidate focusing characteristic analysis model in the candidate data segment estimation network based on the third learning cost index to form a corresponding optimized data segment estimation network.
9. An artificial intelligence based sensor operation behavior mining system comprising a processor and a memory, the memory for storing a computer program, the processor for executing the computer program to implement the method of any one of claims 1-8.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310719547.6A CN116521761B (en) | 2023-06-16 | 2023-06-16 | Sensor operation behavior mining method and system based on artificial intelligence |
CN202311141865.5A CN116955459A (en) | 2023-06-16 | 2023-06-16 | Sensor operation behavior estimation network optimization method and system based on artificial intelligence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310719547.6A CN116521761B (en) | 2023-06-16 | 2023-06-16 | Sensor operation behavior mining method and system based on artificial intelligence |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311141865.5A Division CN116955459A (en) | 2023-06-16 | 2023-06-16 | Sensor operation behavior estimation network optimization method and system based on artificial intelligence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116521761A CN116521761A (en) | 2023-08-01 |
CN116521761B true CN116521761B (en) | 2023-09-29 |
Family
ID=87401377
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311141865.5A Withdrawn CN116955459A (en) | 2023-06-16 | 2023-06-16 | Sensor operation behavior estimation network optimization method and system based on artificial intelligence |
CN202310719547.6A Active CN116521761B (en) | 2023-06-16 | 2023-06-16 | Sensor operation behavior mining method and system based on artificial intelligence |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311141865.5A Withdrawn CN116955459A (en) | 2023-06-16 | 2023-06-16 | Sensor operation behavior estimation network optimization method and system based on artificial intelligence |
Country Status (1)
Country | Link |
---|---|
CN (2) | CN116955459A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107977379A (en) * | 2016-10-25 | 2018-05-01 | 百度国际科技(深圳)有限公司 | Method and apparatus for mined information |
CN107992746A (en) * | 2017-12-14 | 2018-05-04 | 华中师范大学 | Malicious act method for digging and device |
CN111460130A (en) * | 2020-03-27 | 2020-07-28 | 咪咕数字传媒有限公司 | Information recommendation method, device, equipment and readable storage medium |
CN111708785A (en) * | 2020-06-27 | 2020-09-25 | 广州云莫凡信息科技有限公司 | Big data mining processing method and device and cloud platform |
CN113535815A (en) * | 2021-07-08 | 2021-10-22 | 广州康乾信息科技有限公司 | Business operation behavior big data mining method and system suitable for electronic commerce |
CN114757721A (en) * | 2022-05-25 | 2022-07-15 | 淄博至诚电子商务有限公司 | Service prediction analysis method and AI (Artificial Intelligence) mining system for joint big data mining |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10892055B2 (en) * | 2016-08-24 | 2021-01-12 | Panasonic Intellectual Property Management Co., Ltd. | Motor function estimation information generation apparatus, motor function estimation system, motor function estimation information generation method, motor function estimation method and storage medium |
-
2023
- 2023-06-16 CN CN202311141865.5A patent/CN116955459A/en not_active Withdrawn
- 2023-06-16 CN CN202310719547.6A patent/CN116521761B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107977379A (en) * | 2016-10-25 | 2018-05-01 | 百度国际科技(深圳)有限公司 | Method and apparatus for mined information |
CN107992746A (en) * | 2017-12-14 | 2018-05-04 | 华中师范大学 | Malicious act method for digging and device |
CN111460130A (en) * | 2020-03-27 | 2020-07-28 | 咪咕数字传媒有限公司 | Information recommendation method, device, equipment and readable storage medium |
CN111708785A (en) * | 2020-06-27 | 2020-09-25 | 广州云莫凡信息科技有限公司 | Big data mining processing method and device and cloud platform |
CN113535815A (en) * | 2021-07-08 | 2021-10-22 | 广州康乾信息科技有限公司 | Business operation behavior big data mining method and system suitable for electronic commerce |
CN114757721A (en) * | 2022-05-25 | 2022-07-15 | 淄博至诚电子商务有限公司 | Service prediction analysis method and AI (Artificial Intelligence) mining system for joint big data mining |
Also Published As
Publication number | Publication date |
---|---|
CN116955459A (en) | 2023-10-27 |
CN116521761A (en) | 2023-08-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110824587B (en) | Image prediction method, image prediction device, computer equipment and storage medium | |
CN116126945B (en) | Sensor running state analysis method and system based on data analysis | |
CN110888911A (en) | Sample data processing method and device, computer equipment and storage medium | |
CN113011172B (en) | Text processing method, device, computer equipment and storage medium | |
CN112016318A (en) | Triage information recommendation method, device, equipment and medium based on interpretation model | |
CN116304341A (en) | Fraud discrimination method and system based on user network big data | |
CN111666393A (en) | Verification method and device of intelligent question-answering system, computer equipment and storage medium | |
CN113449012A (en) | Internet service mining method based on big data prediction and big data prediction system | |
CN113963205A (en) | Classification model training method, device, equipment and medium based on feature fusion | |
CN115062779A (en) | Event prediction method and device based on dynamic knowledge graph | |
CN115983497A (en) | Time sequence data prediction method and device, computer equipment and storage medium | |
CN116821661A (en) | Time sequence data monitoring method and device, electronic equipment and nonvolatile storage medium | |
CN116501972B (en) | Content pushing method and AI intelligent pushing system based on big data online service | |
CN116048944A (en) | Load trend prediction method and system for big data platform | |
CN115757900B (en) | User demand analysis method and system applying artificial intelligent model | |
CN116521761B (en) | Sensor operation behavior mining method and system based on artificial intelligence | |
CN116720119A (en) | Big data identification method and system applied to multi-terminal service interaction | |
CN116628138A (en) | Logistics order text mining method and system applied to deep learning | |
CN116257632A (en) | Unknown target position detection method and device based on graph comparison learning | |
CN115328753A (en) | Fault prediction method and device, electronic equipment and storage medium | |
CN115222112A (en) | Behavior prediction method, behavior prediction model generation method and electronic equipment | |
CN113822135A (en) | Video processing method, device and equipment based on artificial intelligence and storage medium | |
CN114238822A (en) | Identification method and device of Web service equipment | |
CN114117037A (en) | Intention recognition method, device, equipment and storage medium | |
CN113486142A (en) | Semantic-based word semantic prediction method and computer equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |