CN113569955A - Model training method, user portrait generation method, device and equipment - Google Patents
Model training method, user portrait generation method, device and equipment Download PDFInfo
- Publication number
- CN113569955A CN113569955A CN202110862719.6A CN202110862719A CN113569955A CN 113569955 A CN113569955 A CN 113569955A CN 202110862719 A CN202110862719 A CN 202110862719A CN 113569955 A CN113569955 A CN 113569955A
- Authority
- CN
- China
- Prior art keywords
- label
- sample
- data
- user
- tag
- 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
- G06F18/21—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
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- 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
- G06F18/24—Classification techniques
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (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)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the specification provides a model training method, a user portrait generation device and equipment, and can be applied to the technical field of artificial intelligence. The method comprises the following steps: acquiring sample characteristic data and sample label data; screening complete label data of a first sample user and missing label data of a second sample user from the sample label data; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data of a first sample user and second disturbance characteristic data of a second sample user; determining a first prediction label of the first disturbance characteristic data and a second prediction label of the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first prediction label, the second prediction label and the sample label data; and optimizing the label prediction model based on the loss function. The method ensures the completion of the user label and improves the accuracy of the user portrait.
Description
Technical Field
The embodiment of the specification relates to the technical field of artificial intelligence, in particular to a model training method, a user portrait generation device and equipment.
Background
With the development of artificial intelligence technology in recent years, more and more functions can be realized based on corresponding artificial intelligence models, so that users are helped to process services, the work processing effect is improved, and the service processing efficiency is improved. The user portrait can be realized according to the information obtained by acquiring the corresponding information of the user. Specifically, the user portrait may include various user tags, so that the user tags are used for identification and classification, and operations such as service recommendation and service auxiliary processing may be implemented based on the user portrait in a subsequent service processing stage.
Currently, when a user is portrayed, different user tags are generally generated according to the feature data of the user through a corresponding model. However, with the increase of the complexity of the user data, the model cannot adapt to the change situation of the data, so that when the user tags are determined based on the current model, part of users may have tag loss, that is, the model cannot determine the coincidence situation of the user corresponding to all preset tags according to the current data and the judgment rule, thereby interfering with the generation and utilization of the user portrait in the subsequent process.
For the situation of label missing, when the current model cannot determine the corresponding label according to the data, the missing data can only be fed back to the manager for manual judgment, and under the situation of large data volume, much time and resources are consumed, and the development of the corresponding service is seriously affected. Therefore, a method for accurately and effectively completing the user label is needed.
Disclosure of Invention
An object of the embodiments of the present specification is to provide a model training method, a user portrait generation method, an apparatus and a device, so as to solve a problem of how to supplement missing user tags to improve user portrait accuracy.
In order to solve the above technical problem, an embodiment of the present specification provides a model training method, including: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function.
An embodiment of this specification further provides a model training device, including: the data acquisition module is used for acquiring sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; the data screening module is used for screening complete label data corresponding to a first sample user and missing label data corresponding to a second sample user from the sample label data; the complete tag data comprises data labeled for all preset tag types; the label prediction model training module is used for training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; the data perturbation module is used for perturbing the sample characteristic data to obtain first perturbation characteristic data corresponding to a first sample user and second perturbation characteristic data corresponding to a second sample user; the label determination module is used for respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by utilizing a label prediction model; a loss function construction module for constructing a loss function based on the first prediction label, the second prediction label and the sample label data; and the label prediction model optimization module is used for optimizing the label prediction model based on the loss function.
The embodiment of the present specification further provides a model training device, which includes a memory and a processor; the memory to store computer program instructions; the processor to execute the computer program instructions to implement the steps of: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function.
In order to solve the above technical problem, an embodiment of the present specification further provides a user portrait generation method, including: acquiring target characteristic data and target label data of a target user; the target label data corresponds to at least one target label type; determining a missing tag type based on the target tag data; the missing tag types comprise types of target tag types which are missing compared with preset tag types; inputting the target characteristic data and the target label data into a label prediction model to obtain a prediction label; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function; constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
An embodiment of the present specification further provides a user representation generating apparatus, including: the data acquisition module is used for acquiring target characteristic data and target label data of a target user; the target label data corresponds to at least one target label type; a missing tag type determination module for determining a missing tag type based on the target tag data; the missing tag types comprise types of target tag types which are missing compared with preset tag types; the predicted tag obtaining module is used for inputting the target characteristic data and the target tag data into the tag prediction model to obtain a predicted tag; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function; a target user representation construction module for constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
An embodiment of the present specification also provides a user representation generation device, including a memory and a processor; the memory to store computer program instructions; the processor to execute the computer program instructions to implement the steps of: acquiring target characteristic data and target label data of a target user; the target label data corresponds to at least one target label type; determining a missing tag type based on the target tag data; the missing tag types comprise types of target tag types which are missing compared with preset tag types; inputting the target characteristic data and the target label data into a label prediction model to obtain a prediction label; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function; constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
According to the technical scheme provided by the embodiment of the specification, the embodiment of the specification trains the label prediction model by obtaining the sample characteristic data and the sample label data of the sample user and screening out complete label data in the sample characteristic data, then carries out label prediction on the disturbed characteristic data by disturbing the characteristic data and using the label prediction model, constructs a loss function by combining the predicted label and the original sample label data, optimizes the label prediction model based on the loss function, and then can complement the missing label by using the label prediction model, thereby carrying out user imaging by using the complemented label data. By the method, training and optimization of the corresponding model are realized aiming at the condition of user label data loss, and completion of the user label is ensured, so that the accuracy of user portrait is improved.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the specification, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of a model training method according to an embodiment of the present disclosure;
FIG. 2 is a diagram illustrating a semi-supervised learning process according to an embodiment of the present disclosure;
FIG. 3 is a diagram illustrating an application scenario of a model according to an embodiment of the present disclosure;
FIG. 4 is a flow diagram of a method for user representation generation in accordance with an embodiment of the present disclosure;
FIG. 5 is a block diagram of a model training apparatus according to an embodiment of the present disclosure;
FIG. 6 is a block diagram of a user representation generation apparatus according to an embodiment of the present disclosure;
FIG. 7 is a block diagram of a model training apparatus according to an embodiment of the present disclosure;
FIG. 8 is a block diagram of a user representation generation apparatus according to an embodiment of the present disclosure.
Detailed Description
The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure, and it is obvious that the described embodiments are only a part of the embodiments of the present disclosure, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort shall fall within the protection scope of the present specification.
In order to solve the above technical problem, a model training method according to an embodiment of the present disclosure is introduced. The execution subject of the model training method is model training equipment, and the model training equipment comprises but is not limited to a server, an industrial personal computer, a Personal Computer (PC) and the like. As shown in fig. 1, the model training method may include the following implementation steps.
S110: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type.
The sample user may be a user who has certain sample data to train the model. For example, the sample user may be a user who has a record in the corresponding data system, so that the corresponding data of the user can be called to realize the training of the model in the subsequent step.
The sample feature data may be data for describing features of the user, for example, feature data such as user identity information, business processing records, user evaluation information, and the like, and other types of data may also be set as the sample feature data in practical applications.
In some embodiments, to facilitate utilization of the sample feature data, the sample feature data may be converted into a form of feature vectors, that is, each feature vector is constructed for different feature data to obtain a corresponding feature vector. Specifically, let the feature vector x of each customeri＝[xi,1,…,xi,N]I 1, …, M, full customer feature matrix X ∈ RM*N。
The sample tag data is data for labeling the type of the user in the form of a tag. In this embodiment, at least one preset tag type may be preset, and the preset tag type is used to determine different tags. Accordingly, the sample tag data may be used to identify whether the preset tag types all conform to each other, for example, in a case that the preset tag types include an a tag and a B tag, the sample tag data may be "yes" for the a tag identification and "no" for the B tag identification, so as to implement labeling on the user.
In some real-time approaches, the sample tag data may also be in the form of a vector. In particular, each customer label vector yi＝{yi,1,…,yi,D}. All customer label matrix Y ∈ RM*DWhen a customer has a certain tag d, y i,d1, otherwisei,d0, if the tag is missing, yi,dNull. The label of null is the follow-up requirementThe label to be generated.
Preferably, the lengths of the feature vector and the tag vector may be fixed values in order to facilitate calculation in a subsequent process.
S120: screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data includes data labeled for all preset tag types.
Since the technical purpose of the embodiments of the present specification is to supplement a missing tag, there is a missing tag in the sample tag data of a part of sample users, that is, the tagging is not performed for all preset tag types.
Therefore, in this step, the sample tag data may be screened for complete tag data and missing tag data by the integrity of the tag data. The complete tag data may be data labeled for all preset tag types, the missing tag data is data labeled for only a part of the preset tag types, and the other preset tag types are not labeled.
In some embodiments, while the first sample user and the second sample user are filtered, clustering may be performed on the sample users to obtain at least one user cluster. The data of the users in the user cluster may be relatively similar, that is, the users belong to the same type of user.
The specific clustering mode can be realized by a corresponding clustering model, and the clustering model can be set based on the requirements of practical application, which is not described herein again.
After the user clustering clusters are determined, the mean value and the standard deviation of the sample characteristic data of each user clustering cluster can be respectively calculated, and the Bernoulli distribution of the sample label data of each user clustering cluster can also be calculated. The Bernoulli distribution aims to obtain the probability distribution of the client labels in different clustering clusters. The specific calculation process is set according to the actual application, and is not described herein again.
Based on the above embodiment, preferably, before the user cluster is obtained by clustering, the user information of the sample users may be sent to a manager terminal, and the manager classifies the sample users based on a certain classification rule to obtain a corresponding user group. These user groups are used to describe the general classification of sample users. Correspondingly, in the subsequent steps, clustering is performed respectively for each user group to obtain at least one user cluster. Through the pre-classification of the managers, the clustering can be performed more quickly, conveniently and effectively in the step, the consumption of time and computing resources is reduced, and the scheme is favorably performed.
S130: training a label prediction model using the complete label data and the sample feature data of the first sample user.
After the first sample user is obtained through screening, the complete label data and the sample characteristic data of the first sample user can be used for training a label prediction model. The label prediction model is used for realizing the prediction of the corresponding label according to the characteristic data and the label data of the user so as to complete the completion of the label, or directly realizing the judgment result of the user for the labels of all preset label types. The specific structure and parameters of the label prediction model may be set based on the requirements of the actual application, and are not described herein again.
Specifically, the data of the input model can be set as the customer characteristics Xl∈RM*NThe target variable of the training is Yl∈RM *DObtaining a model of Psupervised。
S140: and disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to the first sample user and second disturbance characteristic data corresponding to the second sample user.
Since only the data of the first sample user is utilized in step S130, and the proportion of the first sample user to all users may not be large, the data amount of the complete label data and the sample feature data corresponding to the first sample user is small, and the model trained based on these data may lack a certain accuracy. Therefore, the data corresponding to the second sample user needs to be utilized into the model training as well.
Specifically, in order to implement utilization of the data of the second sample user, the sample feature data of the first sample user and the sample feature data of the second sample user may be perturbed respectively to obtain the first perturbation feature data and the second perturbation feature data.
In some embodiments, the perturbation may be a random perturbation using a normal distribution to the respective feature data. After the user clustering clusters and the mean value and the standard deviation of the sample characteristic data in each user clustering cluster are determined, the sample characteristic data are disturbed based on the mean value and the standard deviation of the sample characteristic data of each user clustering cluster.
Specifically, the disturbance value is constructed as
In the formula (I), the compound is shown in the specification,
i is the identity of the sample user, j is the identity of a single feature in the feature data,
is the mean value of the sample characteristic data j,
is the standard deviation of the sample feature data j. Reuse formula
Obtaining first perturbation characteristic data and second perturbation characteristic data, wherein,
either the first perturbation characteristic data or the second perturbation characteristic data,
the characteristic data j, j of the sample user i is 1, …, N.
In practical applications, other suitable methods may also be adopted to implement the perturbation on the feature data, and are not limited to the above examples, and are not described herein again.
S150: and respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model.
After the label prediction model is trained by using the data of the first sample user, the label prediction model can predict corresponding labels by using the feature data under the condition of not considering the accuracy, so that the first disturbance feature data and the second disturbance feature data can be input into the label prediction model to respectively obtain corresponding first prediction labels and second prediction labels. The detailed calculation process is not described herein.
S160: constructing a loss function based on the first prediction label, the second prediction label, and the sample label data.
After the first and second prediction labels are obtained, a loss function may be constructed in conjunction with the original sample label data. The loss function is used for evaluating the error of the label prediction model so as to correct and optimize the label prediction model, and further improve the accuracy of the model.
In some embodiments, because the sample tag data includes missing tag data, in order to avoid the missing tag from affecting the construction of the loss function in this step, when the bernoulli distribution of the missing tag data is calculated in the foregoing step, the bernoulli distribution of the missing tag data may be used to complement the type of the default tag missing by the second sample user, and then the loss function is constructed based on the complemented sample tag data, the first predicted tag, and the second predicted tag, so as to improve the accuracy of the calculation result.
In some embodiments, before constructing the loss function, dropout processing may be performed on the tag prediction model to obtain an adjusted prediction model. dropout processing is mainly directed at the generalization capability of the model to increase the robustness of the model. Then, the sample characteristic data of the second sample user can be input into the adjustment prediction model to obtain an adjustment prediction label, then the weighted average is obtained according to the sample label data and the prediction label to obtain a comprehensive prediction label, and finally a loss function is constructed by combining the comprehensive prediction label and the second prediction label.
Based on the above embodiment, when constructing the loss function, the comprehensive prediction tag and the second prediction tag may be combined to construct a tag loss function, then the characteristic loss function may be constructed based on the sample characteristic data and the disturbance characteristic data, and finally the loss function may be constructed based on the tag loss function and the characteristic loss function.
A specific example will be described with reference to fig. 2. Let dropout policy be dropout _ 1. Let the neural network structure of the model be dnn _1, and output the prediction result
Loss function LosslUsing cross entropy, for ylAnd
and (6) solving cross entropy. For customer data sets with incomplete labels, Xu∈RM*N，Yu∈RM*DFor each corresponding one as well
i＝1,…,MuAnd randomly perturbing each characteristic variable of the normal distribution. μ of the normal distribution is the mean of each characteristic variable in each cluster, and σ is the standard deviation of each characteristic variable in each cluster. Let the disturbance value be
j is 1, …, N. I.e. after disturbance
Also for each corresponding
i＝1,…,MuIf, if
If the value is present, namely the label is not missing, the original value is reserved; if it is
In null, the Bernoulli distribution calculated in the previous step is used for randomly generating the value of null to obtain
Will be disturbed
As input data, based on PsupervisedInput data for retraining. Here PsupervisedIn the retraining process, dropout processing is carried out on the neural network structure, and a dropout strategy is set to be dropout _ 2. Let the neural network structure of the model be dnn _2, and output the prediction result
Will be provided with
And
calculating a weighted average to obtain
Alpha is a hyperparameter. At the same time will
The input data is input into the dnn _1 structure in the previous step, and the output prediction result is
Loss function LossuUsing MSE, pair
And
and (6) calculating MSE. The loss function of the final objective solution is
Is a hyper-parameter.
S170: optimizing the label prediction model based on the loss function.
After the loss function is obtained, since the loss function realizes the estimation of the prediction effect of the label prediction model, the prediction of the label prediction model can be realized by combining the loss function. Specifically, model optimization may be implemented by combining a gradient descent method or the like. In practical application, a specific optimization process can be set according to corresponding requirements, and details are not described herein.
A specific scenario example is described below with reference to FIG. 3, which is divided into a client representation model training device and a client representation generating device as shown in FIG. 3. In the client portrait model training device, after clustering client labels, supervised learning is performed by using the client portrait labels, and then unsupervised learning is performed by combining missing label data. And then, after the stock customer portrait is generated through the original model, the portrait is complemented by using the customer portrait model obtained through training, and the online customer portrait generation is completed, so that the customer portrait is accurately and effectively realized, and the business recommendation, the business processing and the like are realized by combining the customer portrait in the subsequent steps.
Based on the introduction of the embodiment and the scene example, it can be seen that in the method, the sample characteristic data and the sample label data of the sample user are obtained, the complete label data in the sample characteristic data is screened out, the label prediction model is trained, then the characteristic data is disturbed, the label prediction model is used for performing label prediction on the disturbed characteristic data, the loss function is built by combining the predicted label and the original sample label data, the label prediction model is optimized based on the loss function, the label prediction model can be used for completing the missing label, and therefore the user image is drawn by using the completed label data. By the method, training and optimization of the corresponding model are realized aiming at the condition of user label data loss, and completion of the user label is ensured, so that the accuracy of user portrait is improved.
A user portrait generation method according to an embodiment of the present specification is introduced based on a model training method corresponding to fig. 1. The execution subject of the user portrait generation method is user portrait generation equipment which comprises but is not limited to a server, an industrial personal computer, a PC and the like. The user representation generation device may be the same device as the model training device or a different device. As shown in FIG. 4, the method for generating a user representation may include the following steps.
S410: acquiring target characteristic data and target label data of a target user; the target tag data corresponds to at least one target tag type.
The target user may be a user who needs to perform tag completion. Before step S410 is executed, the target tag data may be output based on other manners, such as a preset tag judgment model, in combination with the target feature data of the target user. The target tag data has a missing type corresponding to all the preset tag types, and therefore completion needs to be performed in subsequent steps.
In some embodiments, sample data including the target feature data and the target label data may also be used for training the label prediction model, so that the accuracy of the label prediction model is further improved through continuous and repeated training of the data.
S420: determining a missing tag type based on the target tag data; the missing tag types include a type in which the target tag type is missing compared to a preset tag type.
After the target tag data is acquired, the target tag data may be compared with a preset tag type, and a type that the target tag type is lacked compared with the preset tag type is determined as a missing tag type. In the subsequent step, the label prediction model can be used to realize the prediction of the label corresponding to the missing label type.
S430: inputting the target characteristic data and the target label data into a label prediction model to obtain a prediction label; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function.
After inputting the target feature data and target tag data into the tag prediction model, the model may determine a corresponding predicted tag of the actual tag type based on the target feature data and target tag data. The specific training process of the label prediction model may refer to the description in the corresponding embodiment of fig. 1, and is not described herein again.
S440: constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
After the predicted tag is obtained, the predicted tag can be combined with original target tag data to complete construction of the target user portrait of the target user. The target user portrait is used for describing the type of the target user, so that the construction of the user portrait can be realized by combining the acquired label, and the type of the target user can be accurately described. The specific process of constructing the portrait by using the tags can be set based on the actual application condition, and is not described herein again.
In some embodiments, when constructing the target user profile, the target feature data may be compared with each of the preset user cluster, where the preset user cluster is often the same type of user. And determining the cluster closest to the target user according to the comparison result as the target user cluster. Because the target user is relatively close to the users in the cluster, the construction of the portrait of the target user can be completed by combining the corresponding labels and characteristics of the users in the target cluster, and the accuracy of the portrait of the target user is further improved.
A model training apparatus according to an embodiment of the present specification is introduced based on a model training method corresponding to fig. 1. The model training device is arranged on the model training equipment. As shown in fig. 5, the model training apparatus includes the following modules.
A data obtaining module 510, configured to obtain sample feature data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type.
A data filtering module 520, configured to filter complete tag data corresponding to a first sample user and missing tag data corresponding to a second sample user from the sample tag data; the complete tag data includes data labeled for all preset tag types.
A label prediction model training module 530, configured to train a label prediction model using the complete label data and the sample feature data of the first sample user.
And the data perturbation module 540 is configured to perturb the sample feature data to obtain first perturbation feature data corresponding to a first sample user and second perturbation feature data corresponding to a second sample user.
A label determination module 550, configured to determine a first prediction label corresponding to the first perturbation characteristic data and a second prediction label corresponding to the second perturbation characteristic data, respectively, by using a label prediction model.
A loss function construction module 560 configured to construct a loss function based on the first prediction label, the second prediction label, and the sample label data.
A label prediction model optimization module 570 for optimizing the label prediction model based on the loss function.
A user portrait creation apparatus according to an embodiment of the present specification is introduced based on a user portrait creation method corresponding to fig. 4. The user portrait generating device is arranged on the user portrait generating equipment. As shown in FIG. 6, the user representation generation apparatus includes the following modules.
A data obtaining module 610, configured to obtain target feature data and target tag data of a target user; the target tag data corresponds to at least one target tag type.
A missing tag type determining module 620, configured to determine a missing tag type based on the target tag data; the missing tag types include a type in which the target tag type is missing compared to a preset tag type.
A predicted tag obtaining module 630, configured to input the target feature data and the target tag data into a tag prediction model to obtain a predicted tag; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function.
A target user representation construction module 640 for constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
Based on the model training method corresponding to fig. 1, an embodiment of the present specification provides a model training apparatus. As shown in FIG. 7, the model training device may include a memory and a processor.
In this embodiment, the memory may be implemented in any suitable manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid state disk, a U disk, or the like. The memory may be used to store computer program instructions.
In this embodiment, the processor may be implemented in any suitable manner. For example, the processor may take the form of, for example, a microprocessor or processor and a computer-readable medium that stores computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, an embedded microcontroller, and so forth. The processor may execute the computer program instructions to perform the steps of: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function.
Based on the user portrait generation method corresponding to fig. 4, an embodiment of the present specification provides a user portrait generation apparatus. As shown in FIG. 8, the user representation generation device may include a memory and a processor.
In this embodiment, the memory may be implemented in any suitable manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid state disk, a U disk, or the like. The memory may be used to store computer program instructions.
In this embodiment, the processor may be implemented in any suitable manner. For example, the processor may take the form of, for example, a microprocessor or processor and a computer-readable medium that stores computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, an embedded microcontroller, and so forth. The processor may execute the computer program instructions to perform the steps of: acquiring target characteristic data and target label data of a target user; the target label data corresponds to at least one target label type; determining a missing tag type based on the target tag data; the missing tag types comprise types of target tag types which are missing compared with preset tag types; inputting the target characteristic data and the target label data into a label prediction model to obtain a prediction label; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function; constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
It should be noted that the model training method, the user portrait generation device and the user portrait generation equipment can be applied to the technical field of artificial intelligence, and can also be applied to other technical fields except the technical field of artificial intelligence, which is not limited to this.
In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Hardware Description Language), traffic, pl (core universal Programming Language), HDCal (jhdware Description Language), lang, Lola, HDL, laspam, hardward Description Language (vhr Description Language), vhal (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.
The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
From the above description of the embodiments, it is clear to those skilled in the art that the present specification can be implemented by software plus the necessary first hardware platform. Based on such understanding, the technical solutions of the present specification may be essentially or partially implemented in the form of software products, which may be stored in a storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and include instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments of the present specification.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The description is operational with numerous first or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.
This description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
While the specification has been described with examples, those skilled in the art will appreciate that there are numerous variations and permutations of the specification that do not depart from the spirit of the specification, and it is intended that the appended claims include such variations and modifications that do not depart from the spirit of the specification.
Claims (14)
1. A method of model training, comprising:
obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type;
screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types;
training a label prediction model by using the complete label data and the sample characteristic data of the first sample user;
disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user;
respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model;
constructing a loss function based on the first predictive label, the second predictive label, and the sample label data;
optimizing the label prediction model based on the loss function.
2. The method of claim 1, wherein prior to training a label prediction model using the first sample user's full label data and sample feature data, further comprising:
clustering the sample users to obtain at least one user cluster;
respectively calculating the mean value and the standard deviation of the sample characteristic data of each user cluster;
correspondingly, the perturbing the sample feature data to obtain first perturbed feature data corresponding to a first sample user and second perturbed feature data corresponding to a second sample user includes:
and disturbing the sample characteristic data based on the mean value and the standard deviation of the sample characteristic data of each user cluster to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user.
3. The method of claim 2, wherein the perturbing the sample feature data based on the mean and standard deviation of the sample feature data of each user cluster to obtain first perturbed feature data corresponding to a first sample user and second perturbed feature data corresponding to a second sample user comprises:
constructing a disturbance value of
In the formula (I), the compound is shown in the specification,
i is the identity of the sample user, j is the identity of a single feature in the feature data,
is the mean value of the sample characteristic data j,
is the standard deviation of the sample characteristic data j;
using formulas
Obtaining first perturbation characteristic data and second perturbation characteristic data, wherein,
either the first perturbation characteristic data or the second perturbation characteristic data,
the characteristic data j, j of the sample user i is 1, …, N.
4. The method of claim 2, wherein prior to clustering the sample users to obtain at least one user cluster, further comprising:
sending the user information of the sample user to a manager terminal;
receiving a user group fed back by a manager terminal; the user group is used for describing classification conditions of sample users;
correspondingly, the clustering the sample users to obtain at least one user cluster, including:
and clustering each user group respectively to obtain at least one user cluster.
5. The method of claim 1, wherein constructing a loss function based on the first prediction label, the second prediction label, and the sample label data comprises:
calculating a bernoulli distribution of the missing tag data;
supplementing the types of the preset labels missing from the second sample user by using the Bernoulli distribution of the missing label data;
and constructing a loss function based on the supplemented sample label data, the first prediction label and the second prediction label.
6. The method of claim 1, wherein constructing a loss function based on the first prediction label, the second prediction label, and the sample label data comprises:
performing dropout processing on the label prediction model to obtain an adjusted prediction model;
inputting sample characteristic data of a second sample user into the adjustment prediction model to obtain an adjustment prediction label;
obtaining a weighted average of the sample label data and the prediction label to obtain a comprehensive prediction label;
and combining the comprehensive prediction label and the second prediction label to construct a loss function.
7. The method of claim 6, wherein said combining said composite predictive tag and said second predictive tag to construct a loss function comprises:
combining the comprehensive prediction label and the second prediction label to construct a label loss function;
constructing a characteristic loss function based on the sample characteristic data and the disturbance characteristic data;
and constructing a loss function based on the label loss function and the characteristic loss function.
8. A model training apparatus, comprising:
the data acquisition module is used for acquiring sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type;
the data screening module is used for screening complete label data corresponding to a first sample user and missing label data corresponding to a second sample user from the sample label data; the complete tag data comprises data labeled for all preset tag types;
the label prediction model training module is used for training a label prediction model by using the complete label data and the sample characteristic data of the first sample user;
the data perturbation module is used for perturbing the sample characteristic data to obtain first perturbation characteristic data corresponding to a first sample user and second perturbation characteristic data corresponding to a second sample user;
the label determination module is used for respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by utilizing a label prediction model;
a loss function construction module for constructing a loss function based on the first prediction label, the second prediction label and the sample label data;
and the label prediction model optimization module is used for optimizing the label prediction model based on the loss function.
9. A model training apparatus comprising a memory and a processor;
the memory to store computer program instructions;
the processor to execute the computer program instructions to implement the steps of: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function.
10. A user representation generation method, comprising:
acquiring target characteristic data and target label data of a target user; the target label data corresponds to at least one target label type;
determining a missing tag type based on the target tag data; the missing tag types comprise types of target tag types which are missing compared with preset tag types;
inputting the target characteristic data and the target label data into a label prediction model to obtain a prediction label; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function;
constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
11. The method of claim 10, wherein prior to inputting the target feature data and the target tag data into the tag prediction model to obtain the predicted tag, further comprising:
and training the label prediction model by using sample data comprising the target characteristic data and the target label data.
12. The method of claim 10, wherein said constructing a target user representation based on said predictive tags comprises:
comparing the target characteristic data with each preset user cluster respectively;
determining a target user cluster corresponding to the target user according to the comparison result;
and combining the target user cluster, and constructing a target user portrait based on the prediction label.
13. A user representation generation apparatus, comprising:
the data acquisition module is used for acquiring target characteristic data and target label data of a target user; the target label data corresponds to at least one target label type;
a missing tag type determination module for determining a missing tag type based on the target tag data; the missing tag types comprise types of target tag types which are missing compared with preset tag types;
the predicted tag obtaining module is used for inputting the target characteristic data and the target tag data into the tag prediction model to obtain a predicted tag; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function;
a target user representation construction module for constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
14. A user representation generation device comprising a memory and a processor;
the memory to store computer program instructions;
the processor to execute the computer program instructions to implement the steps of: acquiring target characteristic data and target label data of a target user; the target label data corresponds to at least one target label type; determining a missing tag type based on the target tag data; the missing tag types comprise types of target tag types which are missing compared with preset tag types; inputting the target characteristic data and the target label data into a label prediction model to obtain a prediction label; the predicted tag comprises a tag corresponding to the missing tag type; the label prediction model is obtained by the following method: obtaining sample characteristic data and sample label data of at least one sample user; the sample feature data is used for describing features of a sample user; the sample label data is used for labeling at least one preset label type; screening the sample label data for complete label data corresponding to a first sample user and missing label data corresponding to a second sample user; the complete tag data comprises data labeled for all preset tag types; training a label prediction model by using the complete label data and the sample characteristic data of the first sample user; disturbing the sample characteristic data to obtain first disturbance characteristic data corresponding to a first sample user and second disturbance characteristic data corresponding to a second sample user; respectively determining a first prediction label corresponding to the first disturbance characteristic data and a second prediction label corresponding to the second disturbance characteristic data by using a label prediction model; constructing a loss function based on the first predictive label, the second predictive label, and the sample label data; optimizing the label prediction model based on the loss function; constructing a target user representation based on the predicted tags; the target user representation is used to describe a type of the target user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110862719.6A CN113569955A (en) | 2021-07-29 | 2021-07-29 | Model training method, user portrait generation method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110862719.6A CN113569955A (en) | 2021-07-29 | 2021-07-29 | Model training method, user portrait generation method, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113569955A true CN113569955A (en) | 2021-10-29 |
Family
ID=78168858
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110862719.6A Pending CN113569955A (en) | 2021-07-29 | 2021-07-29 | Model training method, user portrait generation method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113569955A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114445146A (en) * | 2022-01-30 | 2022-05-06 | 北京火山引擎科技有限公司 | Label filling method and related equipment thereof |
| CN116774575A (en) * | 2023-08-21 | 2023-09-19 | 山东六佳药用辅料股份有限公司 | Viscosity control method and system for dextrin production process |
-
2021
- 2021-07-29 CN CN202110862719.6A patent/CN113569955A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114445146A (en) * | 2022-01-30 | 2022-05-06 | 北京火山引擎科技有限公司 | Label filling method and related equipment thereof |
| CN116774575A (en) * | 2023-08-21 | 2023-09-19 | 山东六佳药用辅料股份有限公司 | Viscosity control method and system for dextrin production process |
| CN116774575B (en) * | 2023-08-21 | 2023-11-03 | 山东六佳药用辅料股份有限公司 | Viscosity control method and system for dextrin production process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113822494B (en) | Risk prediction method, device, equipment and storage medium | |
| CN110069709B (en) | Intention recognition method, device, computer readable medium and electronic equipment | |
| CN111611488B (en) | Information recommendation method and device based on artificial intelligence and electronic equipment | |
| CN108629358B (en) | Object class prediction method and device | |
| CN111428757B (en) | Model training method, abnormal data detection method and device and electronic equipment | |
| CN113569955A (en) | Model training method, user portrait generation method, device and equipment | |
| CN112800053A (en) | Data model generation method, data model calling device, data model equipment and storage medium | |
| CN111783873A (en) | Incremental naive Bayes model-based user portrait method and device | |
| Terziyan et al. | Causality-aware convolutional neural networks for advanced image classification and generation | |
| Wang et al. | Cost-sensitive active learning for incomplete data | |
| CN113837307A (en) | Data similarity calculation method and device, readable medium and electronic equipment | |
| CN111783688B (en) | Remote sensing image scene classification method based on convolutional neural network | |
| CN113704534A (en) | Image processing method and device and computer equipment | |
| CN112269875A (en) | Text classification method and device, electronic equipment and storage medium | |
| CN116861287A (en) | Online semi-supervised classification algorithm based on multi-label evolution Gao Weiwen flow | |
| CN115905293A (en) | Switching method and device of job execution engine | |
| CN115905715A (en) | Internet data analysis method and platform based on big data and artificial intelligence | |
| CN117194966A (en) | Training method and related device for object classification model | |
| CN115730152A (en) | Big data processing method and big data processing system based on user portrait analysis | |
| CN115146653A (en) | Dialogue script construction method, device, equipment and storage medium | |
| CN111506554B (en) | Data labeling method and related device | |
| CN113159213A (en) | Service distribution method, device and equipment | |
| CN114003591A (en) | Commodity data multi-mode cleaning method and device, equipment, medium and product thereof | |
| CN113222141A (en) | Model supervision training method, device and equipment | |
| CN111091198A (en) | Data processing method and device |
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 |