CN112529109A - Unsupervised multi-model-based anomaly detection method and system - Google Patents
Unsupervised multi-model-based anomaly detection method and system Download PDFInfo
- Publication number
- CN112529109A CN112529109A CN202011585830.7A CN202011585830A CN112529109A CN 112529109 A CN112529109 A CN 112529109A CN 202011585830 A CN202011585830 A CN 202011585830A CN 112529109 A CN112529109 A CN 112529109A
- Authority
- CN
- China
- Prior art keywords
- sample
- model
- abnormal
- training
- models
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an anomaly detection method based on an unsupervised multi-model, which comprises the following steps: collecting a large number of samples, and obtaining a training sample and a sample to be detected; constructing k anomaly detection models; training an anomaly detection model by using a training sample in an unsupervised mode; finally obtaining m abnormity detection models, wherein m is an integer which is more than or equal to 1 and less than or equal to k; when the application is actually deployed, the sample to be detected is input into the m abnormity detection models, and whether the sample to be detected is abnormal or abnormal type is judged according to the results of the m models. The method of the invention completely adopts an unsupervised method to train the model without marking data. Even if the abnormal type is unknown, the abnormal type can be automatically obtained finally. The abnormal sample can be accurately detected without abnormal sample data. The number of the models is changed according to the change of the number of the abnormal categories, and the method is good in robustness and strong in expansibility.
Description
Technical Field
The invention relates to the field of data analysis, in particular to an anomaly detection method and system based on an unsupervised multi-model.
Technical Field
With the rapid development of information technology, a large amount of data is generated in each industry, and the data is often required to be monitored and classified so as to solve the problems in each industry. For example, in the electronic transaction industry, it is necessary to identify whether a user account is stolen or not, and whether fraud or other abnormal behaviors exist in the transaction or not. In the industrial manufacturing industry, quality monitoring is required to be performed on produced products, and abnormal products are detected, so that the number of defective products is reduced, and the product quality is improved. However, in practical cases, the abnormal data is few, a large number of data samples are normal samples without abnormality, and the marking cost of the data is large.
In the prior art, CN 107391569B (method, apparatus and device for data type identification, model training and risk identification) provides a method, apparatus and computer device for data type identification and model training. The method comprises the steps of obtaining data to be identified, and detecting whether the data to be identified is first-class data or not by using a preset anomaly detection model; and inputting other data except the first class data identified by the anomaly detection model into the classification model for identification. The method at least needs to obtain two types of data, namely a normal sample and an abnormal sample, and cannot be applied to a scene without the abnormal sample. CN 109871954 a (training sample generation method, anomaly detection method and apparatus) provides a training sample generation method, anomaly detection method and apparatus. The method solves the problem of insufficient training samples by generating the training samples. However, in practical situations, it is difficult to obtain the true sample distribution through the model, and the generated sample has a large error from the actual sample distribution. CN 108563548B (abnormality detection method and apparatus) discloses an abnormality detection method and apparatus. The method comprises the steps of generating a fault request according to a prefabricated rule, and obtaining context data of the fault request to obtain abnormal data. The abnormal data obtained by the prefabrication rule is inconsistent with the real abnormal data distribution in the actual scene, and the accuracy rate is difficult to ensure.
In short, the prior art needs a large number of training samples and a large number of abnormal samples, or needs to obtain the abnormal samples in a generating mode, and has the problem of low accuracy of abnormal detection.
Disclosure of Invention
The invention aims to overcome the defects in the background art, and provides an unsupervised multi-model-based anomaly detection method and system, which can be used for solving the technical problems that the detection precision is not high and a large number of anomaly samples are needed in the prior art.
In order to achieve the technical effects, the invention adopts the following technical scheme:
an unsupervised multi-model based anomaly detection method, the method comprising the steps of:
s1, collecting a large number of samples, and obtaining training samples and samples to be detected;
s2, constructing k abnormity detection models, wherein k is an integer greater than or equal to 1;
s3, training the 1 st abnormity detection model by using the training sample in an unsupervised mode;
s4, inputting the training sample into the trained 1 st abnormity detection model, and deleting the sample which can be correctly predicted by the model from the training sample to obtain the remaining training sample;
s5, training the remaining k-1 abnormal detection models by using the current remaining training samples to finally obtain m abnormal detection models, wherein m is an integer which is more than or equal to 1 and less than or equal to k;
and S6, when the application is actually deployed, inputting the sample to be detected into m abnormal detection models, and judging whether the sample to be detected is abnormal or abnormal types according to the results of the m models.
Further, the training samples in step S1 at least include normal samples.
Further, the value of k in step S2 is determined by practical application scenarios, including but not limited to:
if the number of the abnormal types of the application scene is known, and the training samples comprise normal samples and samples of all the abnormal types, the value of k is equal to the number of the abnormal types plus 1;
and if the number of the abnormal types of the application scenes is unknown, manually selecting a k value according to experience.
Further, the anomaly detection model of step S2 includes, but is not limited to, a GAN model.
Further, the step S5 of training the remaining k-1 anomaly detection models by using the currently remaining training samples to finally obtain m anomaly detection models includes the following steps:
a. for the ith anomaly detection model, training the model in an unsupervised mode by using the currently remaining training samples, wherein i is an integer and is more than or equal to 2 and less than or equal to k;
b. inputting the remaining training samples into the trained ith abnormal detection model, and deleting the samples which can be correctly predicted by the model from the remaining training samples to obtain the remaining training samples;
c. and (c) repeating the steps a and b, if the number of the samples left at present is 0, terminating the training in advance, otherwise, repeating the steps a and b all the time, and finally obtaining m abnormal detection models.
Furthermore, 1 of the m anomaly detection models corresponds to a model of a normal sample, and the other m-1 models correspond to models of m-1 anomaly type samples.
Further, the step S6 specifically includes:
a. for a sample to be detected, respectively inputting m abnormal detection models, wherein each model outputs a score;
b. if the scores of the m models do not exceed the threshold value t, the sample to be detected is an abnormal sample, and the abnormality is a new category abnormality;
c. if the scores of the models exceed the threshold t, selecting the sample class corresponding to the model with the highest score as the class of the sample to be detected; if the model output score corresponding to the normal sample is the highest, judging that the sample to be detected is the normal sample; and if the output score of the model corresponding to the abnormal sample is the highest, judging that the sample to be detected is abnormal, and determining the abnormal class as the abnormal class corresponding to the model.
Wherein the value of t is determined by: for the training sample corresponding to each model, inputting the training sample into the corresponding model, starting from t =0.5, and increasing by 0.5 each time until t = 0.95. Different values of t correspond to different accuracy rates, and the value of t with the highest accuracy rate is taken as the final value of t.
Meanwhile, the invention also discloses an abnormality detection system based on the unsupervised multiple models, which comprises the following components:
the data acquisition module is used for acquiring data to obtain a training sample and a sample to be detected;
the training module is used for training a plurality of anomaly detection models;
and the deployment application module is used for judging whether the sample to be detected is abnormal or abnormal type.
Compared with the prior art, the invention has the following beneficial effects: even if the abnormal type is unknown, the abnormal type can be automatically obtained finally. The abnormal sample can be accurately detected without abnormal sample data. The model is trained by adopting an unsupervised method without marking data.
Drawings
Fig. 1 is a schematic flowchart of an unsupervised multi-model-based anomaly detection method according to an embodiment of the present invention.
Fig. 2 is a flowchart of model training according to an embodiment of the present invention.
Fig. 3 is a flowchart of a deployment application according to an embodiment of the present invention.
Fig. 4 is a schematic structural diagram of an unsupervised multi-model-based anomaly detection system according to a second embodiment of the present invention.
Detailed Description
The invention will be further elucidated and described with reference to the embodiments of the invention described hereinafter.
Example one
As shown in fig. 1, an unsupervised multi-model-based anomaly detection method specifically includes the following steps:
s1, collecting a large number of samples, and obtaining training samples and samples to be detected;
wherein the training samples at least comprise normal samples;
in this embodiment, the specific implementation is as follows: in an actual production and manufacturing scene, an image acquisition system is used for acquiring a large number of product surface sample pictures including normal samples and abnormal samples to obtain training samples. For training samples, labeling is not required. When the device is deployed, an image acquisition system is used for acquiring a sample picture of the surface of a product to obtain a sample to be detected.
S2, constructing k abnormity detection models, wherein k is an integer greater than or equal to 1;
wherein the value of k is determined by practical application scenarios, including but not limited to:
if the number of the abnormal types of the application scene is known, and the training samples comprise normal samples and samples of all the abnormal types, the value of k is equal to the number of the abnormal types plus 1;
if the number of the abnormal types of the application scene is unknown, manually selecting a k value according to experience;
further, the anomaly detection model of step S2 includes, but is not limited to, a GAN model;
in this embodiment, the types of the abnormalities on the surface of the product are known to be 3, the training sample includes a normal sample and 3 types of abnormal sample pictures, and the k value is determined to be 4 at this time, so 4 GAN models are constructed.
S3, training the 1 st abnormity detection model by using the training sample in an unsupervised mode;
specifically, the steps of training the anomaly detection model GAN model are as follows:
constructing a generating network G by using a convolutional neural network, inputting training samples into G, and obtaining a corresponding generating sample by each sample;
constructing a discrimination network D by using a convolutional neural network, and calculating the probability that the real training sample and a generated sample obtained from G belong to the real sample for each real training sample;
and G needs to make D recognize the sample generated by G as true as possible, D needs to recognize the sample generated by G as false as possible, and after the training is finished, the 1 st trained anomaly detection model, namely the 1 st trained GAN model, is obtained.
S4, inputting the training sample into the trained 1 st abnormity detection model, and deleting the sample which can be correctly predicted by the model from the training sample to obtain the remaining training sample;
specifically, all training samples are input into the 1 st trained GAN model, and samples correctly identified as true samples by the GAN model are deleted from all training samples to obtain the remaining training samples.
S5, training the remaining k-1 abnormal detection models by using the current remaining training samples to finally obtain m abnormal detection models, wherein m is an integer which is more than or equal to 1 and less than or equal to k;
further, the training of the remaining k-1 anomaly detection models by using the currently remaining training samples to finally obtain m anomaly detection models includes the following steps:
a. for the ith anomaly detection model, training the model in an unsupervised mode by using the currently remaining training samples, wherein i is an integer and is more than or equal to 2 and less than or equal to k;
b. inputting the remaining training samples into the trained ith abnormal detection model, and deleting the samples which can be correctly predicted by the model from the remaining training samples to obtain the remaining training samples;
c. repeating the steps a and b, if the number of the samples left at present is 0, terminating the training in advance, otherwise, repeating the steps a and b all the time, and finally obtaining m abnormal detection models;
furthermore, 1 of the m anomaly detection models corresponds to a model of a normal sample, and the other m-1 models correspond to models of m-1 anomaly type samples.
In this embodiment, as shown in fig. 2, the remaining 3 models are trained by using the currently remaining training samples, and the specific steps are as follows:
training a 2 nd GAN model in an unsupervised mode;
deleting the sample which can be correctly predicted by the 2 nd GAN model from the training samples to obtain the remaining training samples;
if the number of training samples left at present is 0, the training is terminated, and finally 2 anomaly detection models are obtained, wherein m = 2.
If the number of the training samples left at present is not 0, judging whether the number of the training models at the moment is larger than or equal to k =4 or not, if not, repeating the steps and continuing the training.
In this embodiment, there are 3 kinds of anomalies and the training samples include all categories of anomaly samples and normal samples, and finally, m =4 trained anomaly detection networks, that is, 4 GAN models, are obtained. Among the 4 anomaly detection models, 1 model corresponds to a model of a normal sample, and the other 3 models correspond to models of 3 anomaly type samples.
And S6, when the application is actually deployed, inputting the sample to be detected into m abnormal detection models, and judging whether the sample to be detected is abnormal or abnormal types according to the results of the m models.
Further, the step S6 specifically includes:
a. for a sample to be detected, respectively inputting m abnormal detection models, wherein each model outputs a score;
b. if the scores of the m models do not exceed the threshold value t, the sample to be detected is an abnormal sample, and the abnormality is a new category abnormality;
c. if the scores of the models exceed the threshold t, selecting the sample class corresponding to the model with the highest score as the class of the sample to be detected; if the model output score corresponding to the normal sample is the highest, judging that the sample to be detected is the normal sample; and if the output score of the model corresponding to the abnormal sample is the highest, judging that the sample to be detected is abnormal, and determining the abnormal class as the abnormal class corresponding to the model.
Wherein the value of t is determined by: for the training sample corresponding to each model, inputting the training sample into the corresponding model, starting from t =0.5, and increasing by 0.5 each time until t = 0.95. Different values of t correspond to different accuracy rates, and the value of t with the highest accuracy rate is taken as the final value of t.
In this embodiment, a flowchart of deploying an application is shown in fig. 3, and the steps are as follows:
for 4 trained models, each model corresponds to a batch of training samples, and the training samples are input into the corresponding models, starting from t =0.5 and increasing by 0.5 each time until t = 0.95. Different values of t correspond to different accuracy rates, and the value of t with the highest accuracy rate is taken as the final value of t.
Collecting a product surface picture by an image collection system to obtain a sample to be detected;
inputting a picture to be detected into m =4 GAN models, wherein each model outputs a score;
and if the scores of the 4 models do not exceed the threshold value t, the sample to be detected is an abnormal sample, and the abnormality is a new category abnormality. And if the model score exceeds the threshold value t, taking the sample class corresponding to the model with the highest score as the class of the sample to be detected, specifically, if the GAN model corresponding to the normal sample has the highest score, judging that the sample to be detected is the normal sample. Otherwise, the sample to be detected is an abnormal sample, and the abnormal type is the abnormal type corresponding to the model.
Example two
Fig. 4 is a schematic structural diagram of an unsupervised multi-model-based anomaly detection system according to a second embodiment of the present invention. The method comprises the following steps: the system comprises a data acquisition module, a training module and a deployment application module.
The data acquisition module is used for acquiring data to obtain a training sample and a sample to be detected;
in this embodiment, the specific implementation is as follows: and acquiring a large number of pictures of the product surface samples including normal samples and abnormal samples by using an image acquisition system to obtain training samples. For training samples, labeling is not required. When the device is deployed, an image acquisition system is used for acquiring a sample picture of the surface of a product to obtain a sample to be detected.
The training module is used for training a plurality of anomaly detection models;
in this embodiment, the specific implementation is as follows: constructing k =4 GAN models, and finally training to obtain 4 abnormal detection models, namely 4 GAN models, by using an unsupervised training method.
And the deployment application module is used for judging whether the sample to be detected is abnormal or abnormal type.
In this embodiment, the specific implementation is as follows: for 4 trained models, each model corresponds to a batch of training samples, and the training samples are input into the corresponding models, starting from t =0.5 and increasing by 0.5 each time until t = 0.95. Different values of t correspond to different accuracy rates, and the value of t with the highest accuracy rate is taken as the final value of t. Collecting a product surface picture by an image collection system to obtain a sample to be detected; the picture to be detected is input into m =4 GAN models, each of which will output a score. And if the scores of the 4 models do not exceed the threshold value t, the sample to be detected is an abnormal sample, and the abnormality is a new category abnormality. And if the model score exceeds the threshold value t, taking the sample class corresponding to the model with the highest score as the class of the sample to be detected, specifically, if the GAN model corresponding to the normal sample has the highest score, judging that the sample to be detected is the normal sample. Otherwise, the sample to be detected is an abnormal sample, and the abnormal type is the abnormal type corresponding to the model.
In summary, the method and system for detecting abnormality based on unsupervised multiple models provided by the invention have the beneficial effects that: the model is trained completely by adopting an unsupervised method without marking data. Even if the abnormal type is unknown, the abnormal type can be automatically obtained finally. The abnormal sample can be accurately detected without abnormal sample data. The number of the models is changed according to the change of the number of the abnormal categories, and the method is good in robustness and strong in expansibility.
It will be understood by those skilled in the art that all or part of the processes of the methods of the above embodiments may be implemented by hardware instructions related to a program, and the program may be stored in a computer-readable storage medium, and when executed, may include the processes of the above embodiments of the methods. The storage medium may be a magnetic disk, an optical disk, a Read-only Memory (ROM), a Random Access Memory (RAM), or the like.
It will be understood that the above embodiments are merely exemplary embodiments taken to illustrate the principles of the present invention, which is not limited thereto. It will be apparent to those skilled in the art that various modifications and improvements can be made without departing from the spirit and substance of the invention, and these modifications and improvements are also considered to be within the scope of the invention.
Claims (8)
1. An unsupervised multi-model-based anomaly detection method is characterized by comprising the following steps of:
s1, collecting a large number of samples, and obtaining training samples and samples to be detected;
s2, constructing k abnormity detection models, wherein k is an integer greater than or equal to 1;
s3, training the 1 st abnormity detection model by using the training sample in an unsupervised mode;
s4, inputting the training sample into the trained 1 st abnormity detection model, and deleting the sample which can be correctly predicted by the model from the training sample to obtain the remaining training sample;
s5, training the remaining k-1 abnormal detection models by using the current remaining training samples to finally obtain m abnormal detection models, wherein m is an integer which is more than or equal to 1 and less than or equal to k;
and S6, inputting the sample to be detected into m abnormal detection models, and judging whether the sample to be detected is abnormal or abnormal type according to the results of the m models.
2. The unsupervised multiple-model-based anomaly detection method according to claim 1, wherein the training samples of step S1 at least comprise normal samples.
3. The unsupervised multiple-model-based anomaly detection method according to claim 1, wherein the value of k in step S2 is determined by practical application scenarios including but not limited to:
if the number of the abnormal types of the application scene is known, and the training samples comprise normal samples and samples of all the abnormal types, the value of k is equal to the number of the abnormal types plus 1;
and if the number of the abnormal types of the application scenes is unknown, manually selecting a k value according to experience.
4. The unsupervised multiple-model-based anomaly detection method according to claim 1, wherein said anomaly detection model of step S2 includes but is not limited to GAN model.
5. The unsupervised multiple-model-based anomaly detection method according to claim 1, wherein the step S5 of training the remaining k-1 anomaly detection models by using the currently remaining training samples to obtain m anomaly detection models comprises the following steps:
a. for the ith anomaly detection model, training the model in an unsupervised mode by using the currently remaining training samples, wherein i is an integer and is more than or equal to 2 and less than or equal to k;
b. inputting the remaining training samples into the trained ith abnormal detection model, and deleting the samples which can be correctly predicted by the model from the remaining training samples to obtain the remaining training samples;
c. and (c) repeating the steps a and b, if the number of the samples left at present is 0, terminating the training in advance, otherwise, repeating the steps a and b all the time, and finally obtaining m abnormal detection models.
6. The m anomaly detection models according to claim 1 or 5, wherein 1 of the models corresponds to a model of a normal sample, and the remaining m-1 models correspond to models of m-1 anomaly class samples.
7. The unsupervised multiple-model-based anomaly detection method according to claim 1, wherein the step S6 specifically comprises:
a. for a sample to be detected, respectively inputting m abnormal detection models, wherein each model outputs a score;
b. if the scores of the m models do not exceed the threshold value t, the sample to be detected is an abnormal sample, and the abnormality is a new category abnormality;
c. if the scores of the models exceed the threshold t, selecting the sample class corresponding to the model with the highest score as the class of the sample to be detected; if the model output score corresponding to the normal sample is the highest, judging that the sample to be detected is the normal sample; and if the output score of the model corresponding to the abnormal sample is the highest, judging that the sample to be detected is abnormal, and determining the abnormal class as the abnormal class corresponding to the model.
8. An unsupervised multi-model-based anomaly detection system applied to the unsupervised multi-model-based anomaly detection method of any one of claims 1 to 7, the method comprising:
the data acquisition module is used for acquiring data to obtain a training sample and a sample to be detected;
the training module is used for training a plurality of anomaly detection models;
and the deployment application module is used for judging whether the sample to be detected is abnormal or abnormal type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011585830.7A CN112529109A (en) | 2020-12-29 | 2020-12-29 | Unsupervised multi-model-based anomaly detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011585830.7A CN112529109A (en) | 2020-12-29 | 2020-12-29 | Unsupervised multi-model-based anomaly detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112529109A true CN112529109A (en) | 2021-03-19 |
Family
ID=74976807
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011585830.7A Pending CN112529109A (en) | 2020-12-29 | 2020-12-29 | Unsupervised multi-model-based anomaly detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112529109A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113114673A (en) * | 2021-04-12 | 2021-07-13 | 西北工业大学 | Network intrusion detection method and system based on generation countermeasure network |
| CN113627576A (en) * | 2021-10-08 | 2021-11-09 | 平安科技（深圳）有限公司 | Code scanning information detection method, device, equipment and storage medium |
| CN114880384A (en) * | 2022-07-11 | 2022-08-09 | 杭州宇谷科技有限公司 | Unsupervised two-wheeled electric vehicle charging time sequence abnormity detection method and system |
| WO2022252079A1 (en) * | 2021-05-31 | 2022-12-08 | 京东方科技集团股份有限公司 | Data processing method and apparatus |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101561878A (en) * | 2009-05-31 | 2009-10-21 | 河海大学 | Unsupervised anomaly detection method and system based on improved CURE clustering algorithm |
| CN107391569A (en) * | 2017-06-16 | 2017-11-24 | 阿里巴巴集团控股有限公司 | Identification, model training, Risk Identification Method, device and the equipment of data type |
| CN107798235A (en) * | 2017-10-30 | 2018-03-13 | 清华大学 | Unsupervised abnormal access detection method and device based on one hot encoding mechanisms |
| CN109447263A (en) * | 2018-11-07 | 2019-03-08 | 任元 | A kind of space flight accident detection method based on generation confrontation network |
| CN111105032A (en) * | 2019-11-28 | 2020-05-05 | 华南师范大学 | Chromosome structure abnormality detection method, system and storage medium based on GAN |
| CN111241673A (en) * | 2020-01-07 | 2020-06-05 | 北京航空航天大学 | Health state prediction method for industrial equipment in noisy environment |
| CN111340791A (en) * | 2020-03-02 | 2020-06-26 | 浙江浙能技术研究院有限公司 | Photovoltaic module unsupervised defect detection method based on GAN improved algorithm |
-
2020
- 2020-12-29 CN CN202011585830.7A patent/CN112529109A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101561878A (en) * | 2009-05-31 | 2009-10-21 | 河海大学 | Unsupervised anomaly detection method and system based on improved CURE clustering algorithm |
| CN107391569A (en) * | 2017-06-16 | 2017-11-24 | 阿里巴巴集团控股有限公司 | Identification, model training, Risk Identification Method, device and the equipment of data type |
| CN107798235A (en) * | 2017-10-30 | 2018-03-13 | 清华大学 | Unsupervised abnormal access detection method and device based on one hot encoding mechanisms |
| CN109447263A (en) * | 2018-11-07 | 2019-03-08 | 任元 | A kind of space flight accident detection method based on generation confrontation network |
| CN111105032A (en) * | 2019-11-28 | 2020-05-05 | 华南师范大学 | Chromosome structure abnormality detection method, system and storage medium based on GAN |
| CN111241673A (en) * | 2020-01-07 | 2020-06-05 | 北京航空航天大学 | Health state prediction method for industrial equipment in noisy environment |
| CN111340791A (en) * | 2020-03-02 | 2020-06-26 | 浙江浙能技术研究院有限公司 | Photovoltaic module unsupervised defect detection method based on GAN improved algorithm |
Non-Patent Citations (4)
| Title |
|---|
| GOGOI P等: "Anomaly detection analysis of intrusion data using supervised & unsupervised approach", 《J. CONVERGENCE INF. TECHNOL》 * |
| H. LIU 等: "Unsupervised multi-target trajectory detection, learning and analysis in complicated environments", 《PROCEEDINGS OF THE 21ST INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR2012)》 * |
| 王祚华等: "改进FCM多分类器组的无监督入侵检测算法", 《小型微型计算机***》 * |
| 陶朝杰等: "基于BalanceCascade-GBDT算法的类别不平衡虚假评论识别方法", 《经济数学》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113114673A (en) * | 2021-04-12 | 2021-07-13 | 西北工业大学 | Network intrusion detection method and system based on generation countermeasure network |
| WO2022252079A1 (en) * | 2021-05-31 | 2022-12-08 | 京东方科技集团股份有限公司 | Data processing method and apparatus |
| CN113627576A (en) * | 2021-10-08 | 2021-11-09 | 平安科技（深圳）有限公司 | Code scanning information detection method, device, equipment and storage medium |
| CN113627576B (en) * | 2021-10-08 | 2022-01-18 | 平安科技（深圳）有限公司 | Code scanning information detection method, device, equipment and storage medium |
| CN114880384A (en) * | 2022-07-11 | 2022-08-09 | 杭州宇谷科技有限公司 | Unsupervised two-wheeled electric vehicle charging time sequence abnormity detection method and system |
| CN114880384B (en) * | 2022-07-11 | 2022-09-23 | 杭州宇谷科技有限公司 | Unsupervised two-wheeled electric vehicle charging time sequence abnormity detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112529109A (en) | Unsupervised multi-model-based anomaly detection method and system | |
| CN111353549B (en) | Image label verification method and device, electronic equipment and storage medium | |
| CN109285791B (en) | Design layout-based rapid online defect diagnosis, classification and sampling method and system | |
| CN112052813B (en) | Method and device for identifying translocation between chromosomes, electronic equipment and readable storage medium | |
| CN112733884A (en) | Welding defect recognition model training method and device and computer terminal | |
| CN113269042B (en) | Intelligent traffic management method and system based on driving vehicle violation identification | |
| CN116453438B (en) | Display screen parameter detection method, device, equipment and storage medium | |
| CN114820598B (en) | PCB defect detection system and PCB defect detection method | |
| CN117152119A (en) | Profile flaw visual detection method based on image processing | |
| CN112348170A (en) | Fault diagnosis method and system for turnout switch machine | |
| CN115601293A (en) | Object detection method and device, electronic equipment and readable storage medium | |
| CN115564776A (en) | Abnormal cell sample detection method and device based on machine learning | |
| CN115620083A (en) | Model training method, face image quality evaluation method, device and medium | |
| CN115494431A (en) | Transformer fault warning method, terminal equipment and computer readable storage medium | |
| WO2022059135A1 (en) | Error cause estimation device and estimation method | |
| CN114897863A (en) | Defect detection method, device and equipment | |
| CN111798237A (en) | Abnormal transaction diagnosis method and system based on application log | |
| CN113239075A (en) | Construction data self-checking method and system | |
| CN118096747B (en) | Automatic PCBA (printed circuit board assembly) board detection method and system based on deep learning | |
| CN111160454B (en) | Quick change signal detection method and device | |
| CN117935174B (en) | Intelligent management system and method for vacuum bag film production line | |
| US20220392187A1 (en) | Image recognition system | |
| CN116894965A (en) | Teacher data collection method and collection device | |
| CN111652323B (en) | Water quality monitoring method, device and server | |
| CN118096747A (en) | Automatic PCBA (printed circuit board assembly) board detection method and system based on deep learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210319 |
|
| RJ01 | Rejection of invention patent application after publication |