CN114118192A - Training method, prediction method, device and storage medium of user prediction model - Google Patents
Training method, prediction method, device and storage medium of user prediction model Download PDFInfo
- Publication number
- CN114118192A CN114118192A CN202010903751.XA CN202010903751A CN114118192A CN 114118192 A CN114118192 A CN 114118192A CN 202010903751 A CN202010903751 A CN 202010903751A CN 114118192 A CN114118192 A CN 114118192A
- Authority
- CN
- China
- Prior art keywords
- user
- feature
- training
- model
- data
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
Abstract
The invention discloses a training method, a prediction method and a prediction device of a user prediction model and a storage medium. The training method comprises the following steps: carrying out data preprocessing on the sample set data; performing feature extraction on the sample set data after the data preprocessing to obtain a feature data set for describing each user; performing feature selection on the feature data set to obtain a target feature set; training a user prediction model based on the target feature set to obtain a trained user prediction model; the user prediction model includes at least two base models and a meta model connecting the at least two base models. The embodiment of the invention can obtain multi-dimensional effective user characteristics; and the user prediction model comprises at least two base models and a meta-model connecting the at least two base models, so that the target feature set of a plurality of users can be predicted by fusing a plurality of base models, and the accuracy of prediction is improved.
Description
Technical Field
The present invention relates to the field of wireless communications, and in particular, to a training method, a prediction method, an apparatus, and a storage medium for a user prediction model.
Background
With the rise of big data, in the related art, a user portrait can be often depicted based on user data, thereby providing powerful support for business development of operators.
For example, 5G (fifth generation mobile communication) users have less data and are often predicted according to preset specified rules, so that feature data are incomplete, the 5G target users are poor in prediction effect, and 5G same-network upgrade users and 5G different-network conversion users cannot be accurately predicted and classified.
Disclosure of Invention
In view of this, embodiments of the present invention provide a training method, a prediction method, an apparatus and a storage medium for a user prediction model, which aim to effectively predict a target user.
The technical scheme of the embodiment of the invention is realized as follows:
the embodiment of the invention provides a training method of a user prediction model, which comprises the following steps:
performing data preprocessing on sample set data, wherein the sample set data comprises user data related to mobile communication of a preset number of users and label information of each user, and the label information is used for representing that the user is one of the following users: target users for upgrading in the same network, target users for upgrading in different networks and non-target users;
performing feature extraction on the sample set data after the data preprocessing to obtain a feature data set for describing each user;
performing feature selection on the feature data set to obtain a target feature set;
training a user prediction model based on the target feature set to obtain a trained user prediction model;
wherein the user prediction model comprises at least two base models and a meta model connecting the at least two base models.
The embodiment of the invention also provides a user prediction method, which comprises the following steps:
inputting the target feature set of the user to be predicted into the user prediction model obtained by training the training method of the user prediction model in the embodiment of the invention, and obtaining the classification result of the user to be predicted.
The embodiment of the present invention further provides a training apparatus for a user prediction model, including:
the system comprises a preprocessing module, a data preprocessing module and a data processing module, wherein the preprocessing module is used for preprocessing sample set data, the sample set data comprises user data related to mobile communication of a preset number of users and label information of each user, and the label information is used for representing that the user is one of the following users: target users for upgrading in the same network, target users for upgrading in different networks and non-target users;
the characteristic extraction module is used for extracting the characteristics of the sample set data after the data preprocessing to obtain a characteristic data set for describing each user;
the characteristic selection module is used for carrying out characteristic selection on the characteristic data set to obtain a target characteristic set;
the model training module is used for training a user prediction model based on the target feature set to obtain a trained user prediction model;
wherein the user prediction model comprises at least two base models and a meta model connecting the at least two base models.
An embodiment of the present invention further provides a user prediction apparatus, including:
and the prediction module is used for inputting the target feature set of the user to be predicted into the user prediction model obtained by training of the training device of the user prediction model in the embodiment of the invention to obtain the classification result of the user to be predicted.
The embodiment of the present invention further provides a training device for a user prediction model, including: a processor and a memory for storing a computer program capable of running on the processor, wherein the processor, when running the computer program, is configured to perform the steps of the training method for the user prediction model according to the embodiment of the present invention.
An embodiment of the present invention further provides a user prediction device, including: a processor and a memory for storing a computer program capable of running on the processor, wherein the processor, when running the computer program, is adapted to perform the steps of the user prediction method of the present invention.
The embodiment of the present invention further provides a storage medium, where a computer program is stored on the storage medium, and when the computer program is executed by a processor, the steps of the method according to any embodiment of the present invention are implemented.
According to the technical scheme provided by the embodiment of the invention, the sample set data after data preprocessing is subjected to feature extraction to obtain a feature data set for describing each user; performing feature selection on the feature data set to obtain a target feature set, so as to obtain multi-dimensional effective user features; the user prediction model of the embodiment of the invention comprises at least two base models and the meta model connected with the at least two base models, so that the target feature set of a plurality of users can be predicted by fusing a plurality of base models, and the accuracy of prediction is improved; in addition, the embodiment of the invention can classify the target users for upgrading in the same network and the target users for upgrading in different networks, enriches the types of the prediction results and is beneficial to the service development of operators.
Drawings
FIG. 1 is a schematic flow chart of a training method of a user prediction model according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart illustrating a process of generating a target feature set based on feature selection in an application example of the present invention;
FIG. 3 is a schematic diagram illustrating a principle of fusing four basis models based on a Stacking fusion method in an application example of the present invention;
FIG. 4 is a flowchart illustrating a user prediction method according to an embodiment of the present invention;
FIG. 5 is a schematic structural diagram of a training apparatus for a user prediction model according to an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a user prediction apparatus according to an embodiment of the present invention;
FIG. 7 is a schematic structural diagram of a training apparatus for a user prediction model according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of a user prediction device according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
The embodiment of the invention provides a training method of a user prediction model, which comprises the following steps of:
illustratively, the target users are 5G users, and the tag information classifies the users into the following three categories: 5G users upgraded in the same network, 5G users upgraded in different networks and non-5G users. The 5G user upgraded on the same network refers to a user upgrading a non-5G user (for example, a 4G or 3G user) to a 5G user under the same operator; the 5G user upgraded by the different network refers to a user upgrading a non-5G user under one operator to a 5G user under another operator.
Here, the sample set data may include mobile communication-related user data collected in different databases, such as data under various carrier platforms, data under application platforms, and the like.
102, performing feature extraction on the sample set data after the data preprocessing to obtain a feature data set for describing each user;
here, the data preprocessing refers to preprocessing for integrating user data from different sources, cleaning the data (including attribute missing value processing, deleting fields which do not make sense for model construction, etc.), data conversion (such as variable category adjustment, attribute conversion, etc.), balancing the data, etc., and obtaining a feature width table matching with a business target, for example, a data table for all features predicted by 5G users.
103, performing feature selection on the feature data set to obtain a target feature set;
and 104, training a user prediction model based on the target feature set to obtain the trained user prediction model.
In an embodiment of the present invention, the user prediction model includes at least two base models and a meta model connecting the at least two base models.
The embodiment of the invention carries out feature extraction on the sample set data after data preprocessing to obtain a feature data set for describing each user; performing feature selection on the feature data set to obtain a target feature set, so as to obtain multi-dimensional effective user features; the user prediction model of the embodiment of the invention comprises at least two base models and the meta model connected with the at least two base models, so that the target feature set of a plurality of users can be predicted by fusing a plurality of base models, and the accuracy of prediction is improved; in addition, the embodiment of the invention can classify the target users for upgrading in the same network and the target users for upgrading in different networks, enriches the types of the prediction results and is beneficial to the service development of operators.
In some embodiments, the feature data set comprises at least one of: the system comprises a user basic attribute feature, a user work information feature, a device use feature, a user consumption preference feature, a user communication feature, a user internet behavior feature and a user social information feature. Illustratively, the feature data set includes all of the features described above, so that potential target users can be fully characterized, and the accuracy of the user prediction model can be improved.
The foregoing features are exemplified in detail below:
the user basic attribute features may include at least one of: age, gender, region, family condition of the user;
the user work information characteristics may include at least one of: whether the user is economically independent, occupation, industry, salary level, working year, month dominable income, work place;
the device usage characteristics may include at least one of: the mobile phone model of the user, whether a plurality of mobile phones exist, whether a plurality of operator mobile phone cards exist, the last mobile phone replacing year, the mobile phone replacing period and whether a 5G mobile phone is used;
the user consumption preference characteristics may include at least one of: monthly consumption type, various money amounts, intelligent equipment consumption types and monthly communication expense;
the user communication characteristics may include at least one of: whether the user is a 4G user, the network access time is 4G, whether the network access time is 4G is a contract package user, monthly flow use condition, package use type, flow package ordering condition, whether the 5G package is used, 5G flow use condition, over-flow times, auxiliary card flow use condition and auxiliary card over-flow times in a fixed time period;
the user internet behavior characteristics may include at least one of: the method comprises the following steps that the online time of a user, the online frequency, the online scene, the Top3 APP consumed by the user traffic, the TOP3 APP used by the user, the frequency and the time for watching high-definition videos, the frequency and the time for watching VR videos, the frequency and the time for playing large-scale mobile phone games, the frequency and the time for watching live broadcasts, the frequency and the time for using remote office software, the frequency and the time for using online courses and the frequency and the time for using online medical treatment are determined;
the user social information feature refers to a location based social networking network (LBSN) -related feature. With the development of smart phones and communication networks, LBSN-related features are used by many people. The sign-in service provided by the LBSN combines the social behaviors and the position information of the user to obtain a large amount of LBSN heterogeneous data including text information, space-time information, audio and video information, picture information and the like, so that the social characteristics of the user are richer. Based on LBSN social data and search data, including data sources such as microblog, WeChat friend circles, search records in webpages and the like, position features and semantic features can be extracted aiming at position information respectively, and related features can include a user sign-in geographical position TOP5 large class, a user sign-in semantic position TOP5 large class, a user sharing 5G related content frequency and a user social sharing topic TOP5 large class.
In order to avoid dimension disasters and reduce the difficulty of model learning tasks, in the embodiment of the invention, truly relevant features are selected from the feature data set, so that a target feature set is constructed. Illustratively, the feature selection may be based on a maximum Mutual Information Coefficient (MIC).
Here, MIC is used to measure the degree of association, linear or non-linear relationship, between two genes with higher accuracy compared to Mutual Information (MI). Wherein a larger MIC value indicates that the two features (dimensions) are closer together. The calculation formula of the MIC value can refer to the prior art, and is not described in detail herein.
In some embodiments, the performing feature selection on the feature data set to obtain a target feature set includes:
adding a first feature with the maximum MIC value between the feature data set and the label information into a target feature set;
traversing other characteristics remained in the characteristic data set, and selecting the characteristic with the maximum mean value of MIC values between the characteristic data set and the label information to be added into the target characteristic set;
and evaluating whether the model training performance of the current target feature set is greater than the model training performance of the last target feature set, if so, continuing to traverse other remaining features in the feature data set, selecting the feature with the maximum mean value of MIC values between the feature data set and the first feature and the label information, adding the feature into the target feature set until the model training performance of the current target feature set is less than or equal to the model training performance of the last target feature set, stopping feature selection, and deleting the feature added last time in the target feature set.
Here, the remaining other features in the feature data set refer to features in the feature data set other than the features that have been added to the target feature set.
Here, the evaluating of the model training performance of the target feature set may be that the target feature set is trained by using a KNN (K-Nearest Neighbor algorithm) model, and indexes such as accuracy and AUC (Area Under Curve) of the model are calculated, so as to evaluate the training performance of the model.
In an application example, as shown in fig. 2, generating the target feature set based on feature selection includes:
here, the original feature data set refers to a feature data set for describing each user, which is obtained by performing feature extraction on sample set data after data preprocessing.
it will be appreciated that this first feature serves as the first feature to be added to the set of target features.
here, a first MIC value between the traversed feature and the first feature and a second MIC value between the tag information may be calculated, respectively, an average value of the first MIC value and the second MIC value may be taken, a feature having a largest average value may be selected, and the selected feature may be added to the target feature set.
for example, the current target feature set and the last target feature set may be trained by using the KNN model, and the accuracy, AUC, and other indexes of the model are calculated, so as to evaluate the performance of each target feature set corresponding to the model training, that is, the first performance of the current target feature set and the second performance of the last target feature set.
at step 206, a final set of target features is generated.
Here, if the first performance is less than or equal to the second performance, it indicates that the added features do not have the effect of optimizing the performance, the feature selection is stopped, and the last added feature in the target feature set is deleted, so as to obtain a final target feature set.
In some embodiments, training a user prediction model based on the target feature set to obtain a trained user prediction model includes:
dividing the sample set data into a training data set and a testing data set based on users, and determining a target feature set of each user in the training data set and a target feature set of each user in the testing data set;
for each base model in the user prediction model, obtaining a first prediction result of the training data set and a second prediction result of the testing data set based on a cross-validation method;
and training the meta-model in the user prediction model based on the first prediction result and the second prediction result of each base model to obtain a trained user prediction model.
In the embodiment of the invention, each base model forms a first-layer learning network, and the meta model forms a second learning network fusing each base model, so that on the basis of extracting multi-dimensional features (namely on the basis of a target feature set), a plurality of weak classification models are fully fused for prediction, and the accuracy of prediction is improved.
In some embodiments, said training said meta-model in said user prediction model based on said first prediction result and said second prediction result of each said base model comprises:
training the meta-model by taking the first prediction result of each base model as a training set to obtain the trained user prediction model;
and taking the second prediction result of each base model as a test set, and evaluating the performance of the trained user prediction model.
Therefore, the second prediction results of the base models can be used as a test set in the embodiment of the invention, so that the performance of the trained user prediction model can be evaluated.
In some embodiments, the obtaining, for each of the base models in the user prediction model, a first prediction result of the training data set and a second prediction result of the testing data set based on a cross-validation method includes:
and dividing the training data set into five parts, and obtaining a first prediction result of the training data set and a second prediction result of the test data set of each base model based on a five-fold cross-validation method.
In an application example, the sample set data may be proportionally divided into a training data set and a test data set, wherein the test data set is 30% in weight. The ratio of 5G users to non-5G users in the training dataset remains around 1: 3. Next, model building and training are performed on the training data set.
In the model construction, the user characteristics of the training data set are used as input, the user conversion behavior (namely the label information of the user) is output, four common machine learning algorithms of a random forest, an xgboost (decision tree), a KNN (K nearest neighbor) and a Bayesian classifier can be adopted to construct four base models, the algorithms of the four base models have larger difference as much as possible, and the learning abilities of different learning models to different characteristics are integrated and effectively combined. And performing model training of a multi-base model on the training data set, wherein a learning neural network is used as a meta-model, the neural network takes the outputs of the four base models as inputs, and a final prediction result is returned.
When the random forest is used for model training, the optimal model can be obtained by adjusting parameters of the number of decision trees and the number of tree nodes. When the decision tree is used for model training, the optimal model is obtained by adjusting parameters such as the depth of the decision tree.
In an application example, four basic models are fused by using a Stacking fusion method, as shown in fig. 3, the four models include: C1-C4, the fusion steps are as follows:
step 1), dividing the training data set into five parts, performing training prediction on one base model in the first layer by using a five-fold cross validation method, splicing 5 times of validation set results into a matrix, obtaining a validation set prediction result (namely a first prediction result) of the base model, and obtaining the first prediction results of other base models in the same way. Illustratively, the first predictors of the four base models are P1, P2, P3, P4, respectively;
and 2) predicting the test data set by the base model trained in the step 1) each time, and performing weighted average on the test data set prediction results of 5 times of the same base model to obtain a test set prediction result (namely a second prediction result) of the base model, and similarly, obtaining first prediction results of other base models. Illustratively, the second predictors of the four base models are T1, T2, T3, T4, respectively;
and 3) paralleling P1, P2, P3 and P4 obtained by 4 basic models into a matrix as a training set, training and learning a second layer model (namely a meta model), and finally learning to obtain a user prediction model P.
And step 4), inputting T1, T2, T3 and T4 obtained by the 4 base models into the user prediction meta model as a test set, so as to obtain a prediction result of test data and evaluate the performance of the user prediction model P.
An embodiment of the present invention further provides a user prediction method, as shown in fig. 4, including:
Here, the user prediction model is obtained by training using the training method of the user prediction model according to the foregoing embodiment of the present invention, and the target feature set may be understood as a target feature set obtained after the user data of the user to be predicted is based on the foregoing data preprocessing, feature extraction, and feature selection. The classification result can support: target users upgraded in the same network, target users upgraded in different networks and non-target users, so that the types of prediction results are enriched, and the service development of operators is facilitated. Illustratively, the target user is a 5G user, and the classification result may be one of the following: 5G users upgraded in the same network, 5G users upgraded in different networks and non-5G users.
In order to implement the method of the embodiment of the present invention, an embodiment of the present invention further provides a training apparatus for a user prediction model, where the training apparatus for the user prediction model corresponds to the training method for the user prediction model, and each step in the training method for the user prediction model is also completely applicable to the training apparatus for the user prediction model.
As shown in fig. 5, the training apparatus for the user prediction model includes: a preprocessing module 501, a feature extraction module 502, a feature selection module 503, and a model training module 504. Wherein the content of the first and second substances,
the preprocessing module 501 is configured to perform data preprocessing on sample set data, where the sample set data includes user data related to mobile communications of a preset number of users and tag information of each user, and the tag information is used to characterize the user as one of the following: target users for upgrading in the same network, target users for upgrading in different networks and non-target users;
the feature extraction module 502 is configured to perform feature extraction on the sample set data after the data preprocessing to obtain a feature data set for describing each user;
the feature selection module 503 is configured to perform feature selection on the feature data set to obtain a target feature set;
the model training module 504 is configured to train a user prediction model based on the target feature set to obtain a trained user prediction model; wherein the user prediction model comprises at least two base models and a meta model connecting the at least two base models.
In some embodiments, model training module 504 is specifically configured to:
dividing the sample set data into a training data set and a testing data set based on users, and determining a target feature set of each user in the training data set and a target feature set of each user in the testing data set;
for each base model in the user prediction model, obtaining a first prediction result of the training data set and a second prediction result of the testing data set based on a cross-validation method;
and training the meta-model in the user prediction model based on the first prediction result and the second prediction result of each base model to obtain a trained user prediction model.
In some embodiments, the model training module 504 trains the meta-models in the user predictive models based on the first and second predictions for each of the base models, including:
training the meta-model by taking the first prediction result of each base model as a training set to obtain the trained user prediction model;
and taking the second prediction result of each base model as a test set, and evaluating the performance of the trained user prediction model.
In some embodiments, model training module 504 obtains a first predicted result of the training data set and a second predicted result of the testing data set based on a cross-validation method for each of the base models in the user prediction model, including:
and dividing the training data set into five parts, and obtaining a first prediction result of the training data set and a second prediction result of the test data set of each base model based on a five-fold cross-validation method.
In some embodiments, the feature data set comprises at least one of: the system comprises a user basic attribute feature, a user work information feature, a device use feature, a user consumption preference feature, a user communication feature, a user internet behavior feature and a user social information feature.
In some embodiments, the feature selection module 503 is specifically configured to:
adding a first feature with the maximum mutual information value coefficient (MIC) value between the feature data set and the label information into a target feature set;
traversing other characteristics remained in the characteristic data set, and selecting the characteristic with the maximum mean value of MIC values between the characteristic data set and the label information to be added into the target characteristic set;
and evaluating whether the model training performance of the current target feature set is greater than the model training performance of the last target feature set, if so, continuing to traverse other remaining features in the feature data set, selecting the feature with the maximum mean value of MIC values between the feature data set and the first feature and the label information, adding the feature into the target feature set until the model training performance of the current target feature set is less than or equal to the model training performance of the last target feature set, stopping feature selection, and deleting the feature added last time in the target feature set.
In practical applications, the preprocessing module 501, the feature extraction module 502, the feature selection module 503, and the model training module 504 may be implemented by a processor in a training apparatus for a user prediction model. Of course, the processor needs to run a computer program in memory to implement its functions.
It should be noted that: in the training apparatus for a user prediction model provided in the above embodiment, when the user prediction model is trained, only the division of the program modules is illustrated, and in practical applications, the processing may be distributed to different program modules according to needs, that is, the internal structure of the apparatus may be divided into different program modules to complete all or part of the processing described above. In addition, the training device of the user prediction model provided in the above embodiments and the training method embodiment of the user prediction model belong to the same concept, and the specific implementation process thereof is described in detail in the method embodiments and is not described herein again.
In order to implement the method according to the embodiment of the present invention, an embodiment of the present invention further provides a user prediction apparatus, where the user prediction apparatus corresponds to the user prediction method, and each step in the user prediction method is also completely applicable to the embodiment of the user prediction apparatus.
As shown in fig. 6, the user prediction apparatus includes: the prediction module 601 is configured to input the target feature set of the user to be predicted into the user prediction model obtained by training the training device of the user prediction model according to the embodiment of the present invention, so as to obtain a classification result of the user to be predicted.
Based on the hardware implementation of the program module, and in order to implement the method according to the embodiment of the present invention, an embodiment of the present invention further provides a training device for a user prediction model. Fig. 7 shows only an exemplary structure of the training apparatus of the user prediction model, not the entire structure, and a part or the entire structure shown in fig. 7 may be implemented as necessary.
As shown in fig. 7, an embodiment of the present invention provides a training apparatus 700 for a user prediction model, including: at least one processor 701, memory 702, user interface 703, and at least one network interface 704. The various components in the training apparatus 700 of the user predictive model are coupled together by a bus system 705. It will be appreciated that the bus system 705 is used to enable communications among the components. The bus system 705 includes a power bus, a control bus, and a status signal bus in addition to a data bus. But for clarity of illustration the various busses are labeled in figure 7 as the bus system 705.
The user interface 703 may include, among other things, a display, a keyboard, a mouse, a trackball, a click wheel, a key, a button, a touch pad, or a touch screen.
The training method of the user prediction model disclosed by the embodiment of the invention can be applied to the processor 701 or realized by the processor 701. The processor 701 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the training method of the user prediction model may be performed by integrated logic circuits of hardware or instructions in the form of software in the processor 701. The Processor 701 may be a general purpose Processor, a Digital Signal Processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. The processor 701 may implement or perform the methods, steps, and logic blocks disclosed in embodiments of the present invention. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the method disclosed by the embodiment of the invention can be directly implemented by a hardware decoding processor, or can be implemented by combining hardware and software modules in the decoding processor. The software module may be located in a storage medium located in the memory 702, and the processor 701 reads information in the memory 702, and completes the steps of the training method of the user prediction model provided in the embodiment of the present invention in combination with hardware thereof.
In an exemplary embodiment, the training Device of the user prediction model may be implemented by one or more Application Specific Integrated Circuits (ASICs), DSPs, Programmable Logic Devices (PLDs), Complex Programmable Logic Devices (CPLDs), FPGAs, general purpose processors, controllers, Micro Controllers (MCUs), microprocessors (microprocessors), or other electronic components, for performing the aforementioned training method of the user prediction model.
Based on the hardware implementation of the program module, and in order to implement the method according to the embodiment of the present invention, an embodiment of the present invention further provides a user prediction device. Fig. 8 shows only an exemplary structure of the user prediction apparatus, not the entire structure, and a part or the entire structure shown in fig. 8 may be implemented as necessary.
As shown in fig. 8, a user prediction apparatus 800 according to an embodiment of the present invention includes: at least one processor 801, memory 802, a user interface 803, and at least one network interface 804. The various components in the user prediction device 800 are coupled together by a bus system 805. It will be appreciated that the bus system 805 is used to enable communications among the components of the connection. The bus system 805 includes a power bus, a control bus, and a status signal bus in addition to a data bus. For clarity of illustration, however, the various buses are labeled as bus system 805 in fig. 8.
The user interface 803 may include, among other things, a display, a keyboard, a mouse, a trackball, a click wheel, a key, a button, a touch pad, or a touch screen.
The memory 802 in embodiments of the present invention is used to store various types of data to support user prediction of device operation. Examples of such data include: any computer program for operating on a user prediction device.
The user prediction method disclosed by the embodiment of the invention can be applied to the processor 801 or implemented by the processor 801. The processor 801 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the user prediction method may be performed by instructions in the form of hardware, integrated logic circuits, or software in the processor 801. The Processor 801 may be a general purpose Processor, a Digital Signal Processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. Processor 801 may implement or perform the methods, steps, and logic blocks disclosed in embodiments of the present invention. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the method disclosed by the embodiment of the invention can be directly implemented by a hardware decoding processor, or can be implemented by combining hardware and software modules in the decoding processor. The software modules may be located in a storage medium located in the memory 802, and the processor 801 reads the information in the memory 802, and performs the steps of the user prediction method provided by the embodiment of the present invention in combination with the hardware thereof.
In an exemplary embodiment, the user prediction apparatus 800 may be implemented by one or more ASICs, DSPs, PLDs, CPLDs, FPGAs, general-purpose processors, controllers, MCUs, microprocessors, or other electronic components for performing the aforementioned user prediction methods.
It will be appreciated that the memories 702, 802 can be either volatile or nonvolatile memory, and can include both volatile and nonvolatile memory. Among them, the nonvolatile Memory may be a Read Only Memory (ROM), a Programmable Read Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a magnetic random access Memory (FRAM), a Flash Memory (Flash Memory), a magnetic surface Memory, an optical disk, or a Compact Disc Read-Only Memory (CD-ROM); the magnetic surface storage may be disk storage or tape storage. Volatile Memory can be Random Access Memory (RAM), which acts as external cache Memory. By way of illustration and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Enhanced Synchronous Dynamic Random Access Memory (Enhanced DRAM), Synchronous Dynamic Random Access Memory (SLDRAM), Direct Memory (DRmb Access), and Random Access Memory (DRAM). The described memory for embodiments of the present invention is intended to comprise, without being limited to, these and any other suitable types of memory.
In an exemplary embodiment, an embodiment of the present invention further provides a storage medium, that is, a computer storage medium, which may specifically be a computer readable storage medium, for example, a memory 702 storing a computer program, where the computer program is executable by a processor 701 of a training apparatus for a user prediction model to complete the steps of the training method for a user prediction model according to the embodiment of the present invention; as another example, a memory 802 is included that stores a computer program that is executable by the processor 801 of the user prediction device to perform the steps described in the user prediction method of the embodiments of the present invention. The computer readable storage medium may be a ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface Memory, optical disk, or CD-ROM, among others.
It should be noted that: "first," "second," and the like are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
In addition, the technical solutions described in the embodiments of the present invention may be arbitrarily combined without conflict.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (12)
1. A method for training a user prediction model, comprising:
performing data preprocessing on sample set data, wherein the sample set data comprises user data related to mobile communication of a preset number of users and label information of each user, and the label information is used for representing that the user is one of the following users: target users for upgrading in the same network, target users for upgrading in different networks and non-target users;
performing feature extraction on the sample set data after the data preprocessing to obtain a feature data set for describing each user;
performing feature selection on the feature data set to obtain a target feature set;
training a user prediction model based on the target feature set to obtain a trained user prediction model;
wherein the user prediction model comprises at least two base models and a meta model connecting the at least two base models.
2. The method of claim 1, wherein training a user prediction model based on the target feature set to obtain a trained user prediction model comprises:
dividing the sample set data into a training data set and a testing data set based on users, and determining a target feature set of each user in the training data set and a target feature set of each user in the testing data set;
for each base model in the user prediction model, obtaining a first prediction result of the training data set and a second prediction result of the testing data set based on a cross-validation method;
and training the meta-model in the user prediction model based on the first prediction result and the second prediction result of each base model to obtain a trained user prediction model.
3. The method of claim 2, wherein training the meta-models in the user prediction model based on the first and second prediction results for each of the base models comprises:
training the meta-model by taking the first prediction result of each base model as a training set to obtain the trained user prediction model;
and taking the second prediction result of each base model as a test set, and evaluating the performance of the trained user prediction model.
4. The method of claim 2, wherein the obtaining a first prediction result of the training data set and a second prediction result of the testing data set based on a cross-validation method for each of the base models in the user prediction model comprises:
and dividing the training data set into five parts, and obtaining a first prediction result of the training data set and a second prediction result of the test data set of each base model based on a five-fold cross-validation method.
5. The method of claim 1, wherein the feature data set comprises at least one of: the system comprises a user basic attribute feature, a user work information feature, a device use feature, a user consumption preference feature, a user communication feature, a user internet behavior feature and a user social information feature.
6. The method of claim 1, wherein the feature selection of the feature data set to obtain a target feature set comprises:
adding a first feature with the maximum mutual information value coefficient (MIC) value between the feature data set and the label information into a target feature set;
traversing other characteristics remained in the characteristic data set, and selecting the characteristic with the maximum mean value of MIC values between the characteristic data set and the label information to be added into the target characteristic set;
and evaluating whether the model training performance of the current target feature set is greater than the model training performance of the last target feature set, if so, continuing to traverse other remaining features in the feature data set, selecting the feature with the maximum mean value of MIC values between the feature data set and the first feature and the label information, adding the feature into the target feature set until the model training performance of the current target feature set is less than or equal to the model training performance of the last target feature set, stopping feature selection, and deleting the feature added last time in the target feature set.
7. A user prediction method, comprising:
inputting the target feature set of the user to be predicted into the user prediction model obtained by training according to the method of any one of claims 1 to 6, and obtaining the classification result of the user to be predicted.
8. An apparatus for training a user prediction model, comprising:
the system comprises a preprocessing module, a data preprocessing module and a data processing module, wherein the preprocessing module is used for preprocessing sample set data, the sample set data comprises user data related to mobile communication of a preset number of users and label information of each user, and the label information is used for representing that the user is one of the following users: target users for upgrading in the same network, target users for upgrading in different networks and non-target users;
the characteristic extraction module is used for extracting the characteristics of the sample set data after the data preprocessing to obtain a characteristic data set for describing each user;
the characteristic selection module is used for carrying out characteristic selection on the characteristic data set to obtain a target characteristic set;
the model training module is used for training a user prediction model based on the target feature set to obtain a trained user prediction model;
wherein the user prediction model comprises at least two base models and a meta model connecting the at least two base models.
9. A user prediction apparatus, comprising:
the prediction module is used for inputting the target feature set of the user to be predicted into the user prediction model obtained by training of the training device of the user prediction model according to claim 8, and obtaining the classification result of the user to be predicted.
10. An apparatus for training a user prediction model, comprising: a processor and a memory for storing a computer program capable of running on the processor, wherein,
the processor, when executing the computer program, is adapted to perform the steps of the method of any of claims 1 to 6.
11. A user prediction device, comprising: a processor and a memory for storing a computer program capable of running on the processor, wherein,
the processor, when executing the computer program, performs the steps of the method of claim 7.
12. A storage medium having a computer program stored thereon, the computer program, when executed by a processor, implementing the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010903751.XA CN114118192A (en) | 2020-09-01 | 2020-09-01 | Training method, prediction method, device and storage medium of user prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010903751.XA CN114118192A (en) | 2020-09-01 | 2020-09-01 | Training method, prediction method, device and storage medium of user prediction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114118192A true CN114118192A (en) | 2022-03-01 |
Family
ID=80360338
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010903751.XA Pending CN114118192A (en) | 2020-09-01 | 2020-09-01 | Training method, prediction method, device and storage medium of user prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114118192A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115439206A (en) * | 2022-11-08 | 2022-12-06 | 税友信息技术有限公司 | Declaration data prediction method, device, equipment and medium |
| CN116258579A (en) * | 2023-04-28 | 2023-06-13 | 成都新希望金融信息有限公司 | Training method of user credit scoring model and user credit scoring method |
| CN116416884A (en) * | 2023-06-12 | 2023-07-11 | 深圳市彤兴电子有限公司 | Testing device and testing method for display module |
-
2020
- 2020-09-01 CN CN202010903751.XA patent/CN114118192A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115439206A (en) * | 2022-11-08 | 2022-12-06 | 税友信息技术有限公司 | Declaration data prediction method, device, equipment and medium |
| CN115439206B (en) * | 2022-11-08 | 2023-03-07 | 税友信息技术有限公司 | Declaration data prediction method, device, equipment and medium |
| CN116258579A (en) * | 2023-04-28 | 2023-06-13 | 成都新希望金融信息有限公司 | Training method of user credit scoring model and user credit scoring method |
| CN116416884A (en) * | 2023-06-12 | 2023-07-11 | 深圳市彤兴电子有限公司 | Testing device and testing method for display module |
| CN116416884B (en) * | 2023-06-12 | 2023-08-18 | 深圳市彤兴电子有限公司 | Testing device and testing method for display module |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111444428B (en) | Information recommendation method and device based on artificial intelligence, electronic equipment and storage medium | |
| US11640494B1 (en) | Systems and methods for construction, maintenance, and improvement of knowledge representations | |
| CN110321482B (en) | Information recommendation method, device and equipment | |
| US20220114199A1 (en) | System and method for information recommendation | |
| US20200226133A1 (en) | Knowledge map building system and method | |
| CN109960761B (en) | Information recommendation method, device, equipment and computer readable storage medium | |
| CN114118192A (en) | Training method, prediction method, device and storage medium of user prediction model | |
| CN111753198A (en) | Information recommendation method and device, electronic equipment and readable storage medium | |
| CN112396108A (en) | Service data evaluation method, device, equipment and computer readable storage medium | |
| CN111538794B (en) | Data fusion method, device and equipment | |
| CN112799747A (en) | Intelligent assistant evaluation and recommendation method, system, terminal and readable storage medium | |
| WO2021155691A1 (en) | User portrait generating method and apparatus, storage medium, and device | |
| CN111984784B (en) | Person post matching method, device, electronic equipment and storage medium | |
| CN111898675B (en) | Credit wind control model generation method and device, scoring card generation method, machine readable medium and equipment | |
| Tu et al. | Bidirectional sensing of user preferences and application changes for dynamic mobile app recommendations | |
| Fischer et al. | The impact of mobile computing on individuals, organizations, and society-synthesis of existing literature and directions for future research | |
| US20230244862A1 (en) | Form processing method and apparatus, device, and storage medium | |
| WO2024021685A1 (en) | Reply content processing method and media content interactive content interaction method | |
| CN112257959A (en) | User risk prediction method and device, electronic equipment and storage medium | |
| CN111784384A (en) | Payment service data processing method, device, equipment and system | |
| CN112015912B (en) | Intelligent index visualization method and device based on knowledge graph | |
| CN111259975B (en) | Method and device for generating classifier and method and device for classifying text | |
| CN115543428A (en) | Simulated data generation method and device based on strategy template | |
| CN111242520B (en) | Feature synthesis model generation method and device and electronic equipment | |
| Motohashi et al. | Technological competitiveness of China's internet platformers: comparison of Google and Baidu by using patent text information |
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 |