CN115146717A

CN115146717A - Recommendation method and device based on multitask model, electronic equipment and storage medium

Info

Publication number: CN115146717A
Application number: CN202210723718.8A
Authority: CN
Inventors: 汪利伟; 仲籽彦; 魏丫丫; 金伟德; 洪迪; 刘健
Original assignee: China Telecom Corp Ltd
Current assignee: China Telecom Corp Ltd
Priority date: 2022-06-24
Filing date: 2022-06-24
Publication date: 2022-10-04

Abstract

The embodiment of the invention provides a recommendation method, a recommendation device, electronic equipment and a storage medium based on a multitask model, which comprise the following steps: acquiring user behavior data from the log data according to the mobile phone number of the user; screening feature data from the user behavior data, wherein the feature data comprises basic features, interactive features and time slice features; adopting the characteristic data to carry out a prediction task aiming at voice use and a prediction task aiming at flow use through a multi-task model to obtain a prediction result; and determining the voice package and the flow package recommended for the user according to the prediction result. In the embodiment of the invention, the two tasks of the flow and the voice can be respectively predicted by using the multi-task model, so that the prediction efficiency is improved, the package of the most suitable user can be more accurately matched, the service is furthest realized, and the user is kept.

Description

Recommendation method and device based on multitask model, electronic equipment and storage medium

Technical Field

The present invention relates to the field of online marketing technologies, and in particular, to a recommendation method based on a multitask model, a recommendation apparatus based on a multitask model, an electronic device, and a computer-readable storage medium.

Background

The package selected by the user during network access is often not matched with the subsequent use habit, so that the viscosity of the user is not high, and the user is lost. Moreover, with the advent of the 5G era, more and more users need to change the mobile phone cards from 4G to 5G, so that more users can be reserved when the users take their numbers to change networks, how to recommend a more suitable package with the users is required, and the trust of the users on products and the viscosity of the users are urgently solved.

Disclosure of Invention

In view of the above, embodiments of the present invention are proposed to provide a multitask model based recommendation method, a multitask model based recommendation apparatus, an electronic device and a computer readable storage medium that overcome or at least partially solve the above problems.

In order to solve the above problems, an embodiment of the present invention discloses a recommendation method based on a multitask model, where the method includes:

acquiring user behavior data from the log data according to the mobile phone number of the user;

screening feature data from the user behavior data, wherein the feature data comprises basic features, interactive features and time slice features;

by adopting the characteristic data, performing a prediction task aiming at voice use and a prediction task aiming at flow use through a multi-task model to obtain a prediction result;

and determining the voice package and the flow package recommended for the user according to the prediction result.

Optionally, the determining, according to the prediction result, a voice package and a traffic package recommended to the user includes:

if the prediction result is the flow and the voice duration, mapping the flow and the voice duration into a voice package and a flow package according to a preset mapping relation;

and combining the voice package and the flow package into a basic package to be recommended to the user.

Optionally, the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; the step of obtaining a prediction result by adopting the characteristic data and performing a prediction task for voice use and a prediction task for flow use through a multi-task model comprises the following steps:

processing the characteristic data through each gate network to obtain the weight of a plurality of expert networks corresponding to each gate network;

respectively extracting the features of the feature data through a plurality of expert networks corresponding to each gate network to obtain a plurality of first features;

and according to the weights of the expert networks corresponding to each gate network, carrying out weighted summation on the first characteristics corresponding to each gate network, and obtaining the prediction result through the Tower network corresponding to each gate network.

Optionally, the processing the feature data through each gate network to obtain the weight of the plurality of expert networks corresponding to each gate network includes:

performing feature extraction on the feature data through each gate network to obtain the probability that a plurality of expert networks are selected by each gate network;

and carrying out normalization processing on the probability of the plurality of expert networks selected by each gate network to obtain the weight of the plurality of expert networks corresponding to each gate network.

Optionally, the multitask model is trained by:

obtaining feature sample data, wherein the feature sample data comprises a basic feature sample, an interactive feature sample and a time slice feature sample;

taking the feature sample data as an input of the multitask model;

in the multitask model, the characteristic sample data is adopted to carry out a prediction task aiming at voice use and a prediction task aiming at flow use so as to obtain a prediction result;

and training the multi-task model according to the prediction result.

Optionally, the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; in the multitask model, performing a prediction task for voice use and a prediction task for traffic use by using the feature sample data to obtain a prediction result, including:

processing the characteristic sample data through each gate network to obtain the weight of a plurality of expert networks corresponding to each gate network;

respectively extracting the features of the feature sample data through a plurality of expert networks corresponding to each gate network to obtain a plurality of first feature samples;

and according to the weights of a plurality of expert networks corresponding to each gate network, carrying out weighted summation on a plurality of first characteristic samples corresponding to each gate network, and obtaining the prediction results of the flow and the voice duration through the Tower network corresponding to each gate network.

Optionally, the processing, by each gate network, the feature sample data to obtain weights of a plurality of expert networks corresponding to each gate network includes:

performing feature extraction on the feature sample data through each gate network to obtain the probability of selecting a plurality of expert networks by each gate network;

The embodiment of the invention also discloses a recommendation device based on the multitask model, which comprises the following steps:

the data acquisition module is used for acquiring user behavior data from the log data according to the mobile phone number of the user;

the screening characteristic module is used for screening characteristic data from the user behavior data, wherein the characteristic data comprises basic characteristics, interactive characteristics and time slice characteristics;

the characteristic prediction module is used for performing a prediction task aiming at voice use and a prediction task aiming at flow use through a multi-task model by adopting the characteristic data to obtain a prediction result;

and the package determining module is used for determining the voice package and the flow package recommended to the user according to the prediction result.

Optionally, the package determining module includes:

the package mapping submodule is used for mapping the flow and the voice duration into a voice package and a flow package according to a preset mapping relation if the prediction result is the flow and the voice duration;

and the package recommending submodule is used for combining a basic package according to the voice package and the flow package so as to recommend the basic package to the user.

Optionally, the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; the feature prediction module includes:

the weight obtaining submodule is used for processing the characteristic data through each gate network so as to obtain the weight of a plurality of expert networks corresponding to each gate network;

the characteristic extraction submodule is used for respectively extracting the characteristics of the characteristic data through a plurality of expert networks corresponding to each gate network so as to obtain a plurality of first characteristics;

and the prediction result sub-module is used for carrying out weighted summation on a plurality of first characteristics corresponding to each gate network according to the weights of a plurality of expert networks corresponding to each gate network, and obtaining the prediction result through the Tower network corresponding to each gate network.

Optionally, the weight obtaining sub-module includes:

an obtaining probability unit, configured to perform feature extraction on the feature data through each gate network to obtain a probability that the plurality of expert networks are selected by each gate network;

and the weight obtaining unit is used for carrying out normalization processing on the probability of the plurality of expert networks selected by each gate network to obtain the weight of the plurality of expert networks corresponding to each gate network.

Optionally, the multitask model is trained by:

the system comprises a sample acquisition module, a time slice characteristic sample acquisition module and a time slice characteristic sample acquisition module, wherein the sample acquisition module is used for acquiring characteristic sample data which comprises a basic characteristic sample, an interactive characteristic sample and a time slice characteristic sample;

the sample input module is used for taking the characteristic sample data as the input of the multitask model;

the sample prediction module is used for performing a prediction task aiming at voice use and a prediction task aiming at flow use by adopting the characteristic sample data in the multitask model so as to obtain a prediction result;

and the model training module trains the multi-task model according to the prediction result.

Optionally, the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; the sample prediction module comprises:

the sample weight submodule is used for processing the characteristic sample data through each gate network so as to obtain the weight of a plurality of expert networks corresponding to each gate network;

the sample extraction submodule is used for respectively extracting the characteristics of the characteristic sample data through a plurality of expert networks corresponding to each gate network so as to obtain a plurality of first characteristic samples;

and the sample prediction submodule is used for carrying out weighted summation on a plurality of first characteristic samples corresponding to each gate network according to the weights of a plurality of expert networks corresponding to each gate network, and obtaining a prediction result of flow and voice duration through a Tower network corresponding to each gate network.

Optionally, the sample weight submodule includes:

a sample probability unit, configured to perform feature extraction on the feature sample data through each gate network to obtain a probability that the plurality of expert networks are selected by each gate network;

and the sample weight unit is used for carrying out normalization processing on the probability of the plurality of expert networks selected by each gate network to obtain the weight of the plurality of expert networks corresponding to each gate network.

The embodiment of the invention has the following advantages:

in the embodiment of the invention, user behavior data is obtained from log data according to the mobile phone number of a user; screening characteristic data from the user behavior data, and performing a prediction task aiming at voice use and a prediction task aiming at flow use through a multi-task model by adopting the characteristic data to obtain a prediction result; and determining the voice package and the flow package recommended for the user according to the prediction result. In the embodiment of the invention, the two tasks of the flow and the voice can be respectively predicted by using the multi-task model, so that the prediction efficiency is improved, the package of the most suitable user can be more accurately matched, the service is furthest realized, and the user is kept.

Drawings

FIG. 1 is a flowchart illustrating steps of a method for multi-task model based recommendation according to an embodiment of the present invention;

FIG. 2 is a block diagram of a multitasking model according to an embodiment of the present invention;

FIG. 3 is a flowchart of a method for training a multitask model according to an embodiment of the present invention;

fig. 4 is a block diagram of a recommendation device based on a multitask model according to an embodiment of the present invention.

Detailed Description

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.

Along with the arrival of the 5G era, more and more users need to replace the mobile phone card into a 5G form, in order to be able to service and improve the number portability conversion rate of the user to the utmost extent, then need the most suitable package of accurate prediction user, but the most suitable package of prediction user is not accurate at present, and prediction efficiency is lower, therefore, for the most suitable package of more accurate prediction user, and improve prediction efficiency, improve better service for the user, thereby improve the number portability conversion rate of the user.

One of the core ideas of the embodiment of the invention is that data processing is carried out through user behavior data when a user uses a package, the user behavior data after data processing is input to a trained multi-task model which carries out prediction based on two tasks, namely flow and voice, for processing, so that a prediction result is obtained, the prediction efficiency is improved, meanwhile, the most appropriate package can be matched for the user more accurately, the user experience is further improved, and the service and the user can be saved to the maximum extent.

Referring to fig. 1, a flowchart illustrating steps of a recommendation method based on a multitask model according to an embodiment of the present invention is shown, where the method specifically includes the following steps:

step 101, acquiring user behavior data from log data according to a user mobile phone number;

illustratively, log data is stored in the log table, the log table records the relevant behavior information of the user, the user behavior data can be obtained from the log table according to the mobile phone number of the user, and the user behavior data can include user basic information, user behavior information, user terminal information and the like. The user basic information may include the user age, the user gender, the current province and the current package type, etc.; the user behavior information can comprise the times of purchasing the refueling packet by the user, the type of purchasing the refueling packet, the clicking behavior of the user, the charge amount in the period of nearly N months, the flow usage amount in the period of nearly N months and the like; the user terminal information may include the number of terminals used by the user, the brand of the terminal, and the like.

102, screening feature data from the user behavior data, wherein the feature data comprises basic features, interactive features and time slice features;

for example, the specific operation of filtering the feature data may be to perform a preliminary selection on the user behavior data, and then perform data cleansing on the preliminarily selected user behavior data.

The preliminary selection of the user behavior data may be to perform visual analysis on the user behavior data, for example, to perform preliminary selection on the user behavior data through a box chart, a thermodynamic diagram, or the like, and to remove user behavior data with too strong correlation or a missing value or a serious abnormal value; the user behavior data is examined using a random forest to filter out extraneous data.

Performing data cleaning on the preliminarily selected user behavior data, for example, converting the format of the data, such as uniformly converting the money class label into a unit of element; carrying out complementary value processing on data with a non-serious missing value by using a median and a mode according to an actual situation; reassigning the abnormal value by using a k-means clustering algorithm; performing id processing on continuous characteristics such as age and the like; and crossing important characteristics such as age, gender and the like, constructing crossed characteristics and the like. Specifically, how to screen out the feature data can be determined by those skilled in the art according to actual situations, and the present invention is not limited herein.

In an embodiment of the present invention, the feature data may include a basic feature, an interactive feature, and a time slice feature; wherein, the basic characteristics can comprise the age, the sex and the like of the user; the interactive characteristics can comprise the behavior of the user on packages or flow packets, package popularity and the like; time slice characteristics may include the previous day, approximately three days, the end of the middle month of the early month of the month, approximately one month, approximately three months, etc.

103, performing a prediction task aiming at voice use and a prediction task aiming at flow use through a multi-task model by adopting the characteristic data to obtain a prediction result;

in the embodiment of the invention, the characteristic data screened by the user behavior data is input into the trained multi-task model to obtain a corresponding prediction result. The trained multi-task model is based on a voice task and a flow task for prediction modeling. Specifically, which way to train the multitask model can be determined according to actual situations, and the invention is not limited herein.

Fig. 2 is a structural diagram of a multitasking model according to an embodiment of the present invention. The multitasking model may include two Gate networks (Gate a and Gate B), a plurality of Expert networks (Expert), and Tower networks corresponding to the Gate networks one to one; each task is configured with a Gate network, the number of the expert networks may be multiple, specifically, the data of the expert networks is determined according to the actual situation, and the present invention is not limited herein.

The door network can process the input characteristic data to obtain the selection probability of the expert network corresponding to the door network, and the characteristic data is weighted and summed based on a plurality of expert networks and output to the corresponding Tower network; the Tower network is used to represent the final output, and the expression is as follows:

wherein, g ^k (x) The output of Gate is represented as a multilayer perceptron model, implemented as a simple linear transformation plus a softmax (flexible maximum transfer function) layer.

softmax is used to assign a probability value to each output categorical result, indicating the likelihood of belonging to each category.

Loss function: the RMSE (Root Mean Square Error) for the two tasks is calculated separately and then summed as a function of the loss for the entire task.

In an alternative embodiment, the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; said step 103 may comprise the following substeps S11-S13:

substep S11, processing the feature data through each gate network to obtain the weight of a plurality of expert networks corresponding to each gate network;

in the embodiment of the invention, the trained multitask model can comprise two gate networks, a plurality of expert networks and Tower networks which are in one-to-one correspondence with the gate networks; in specific implementation, gate a may be used as a voice task, and Gate B may be used as a flow task; the Gate A can process the input feature data so as to obtain the weights of a plurality of expert networks corresponding to the voice tasks; gate B may process the input feature data to obtain weights of multiple expert networks corresponding to the traffic tasks. Specifically, gate a may also be used as a traffic task, gate B may also be used as a voice task, and a Gate network corresponding to each task may be set according to an actual situation, which is not limited herein.

A substep S12, respectively extracting the characteristic data through a plurality of expert networks corresponding to each gate network to obtain a plurality of first characteristics;

and a substep S13 of performing weighted summation on a plurality of first characteristics corresponding to each gate network according to the weights of a plurality of expert networks corresponding to each gate network, and obtaining the prediction result through a Tower network corresponding to each gate network.

In the embodiment of the invention, each gate network can correspond to a plurality of expert networks, and the plurality of expert networks respectively extract the input characteristic data so as to obtain a plurality of corresponding first characteristics; and carrying out weighted summation on a plurality of first characteristics corresponding to each gate network based on the weights of a plurality of expert networks corresponding to each gate network, and obtaining a prediction result through the Tower network corresponding to each gate network. Wherein the number of first features is the same as the number of expert networks.

In an alternative embodiment, the substep S11 may comprise: performing feature extraction on the feature data through each gate network to obtain the probability of selecting a plurality of expert networks by each gate network; and carrying out normalization processing on the probability of the plurality of expert networks selected by each gate network to obtain the weight of the plurality of expert networks corresponding to each gate network.

In the embodiment of the invention, each gate network can perform feature extraction on the input feature data to obtain a second feature of the input feature data; and performing softmax processing on the second characteristic through the gate network to obtain the probability of the plurality of expert networks selected by the gate network, and performing normalization processing on the probability of the plurality of expert networks corresponding to each gate network to obtain the weight of the plurality of expert networks corresponding to each gate network. Wherein the probability may be 0 in magnitude.

And step 104, determining the voice package and the flow package recommended for the user according to the prediction result.

Illustratively, different Tower networks are set based on different task prediction requirements, for example, if the click rate of the user is predicted, a Tower network for predicting the click rate of the user can be set, and the click probability of the user is obtained through the Tower network.

In the embodiment of the invention, a prediction task aiming at voice use and a prediction task aiming at flow use can be carried out through a trained multi-task model, and a prediction result is obtained through a set Tower network; wherein, the prediction result can be adjusted by setting the Tower network.

In an alternative embodiment, the step 104 may comprise the following sub-steps S21-S22:

step S21, if the prediction result is the flow and the voice duration, mapping the flow and the voice duration into a voice package and a flow package according to a preset mapping relation;

and a substep S22, combining a basic package according to the voice package and the flow package so as to recommend the basic package to the user.

In the embodiment of the invention, the prediction result can be flow and voice duration; and mapping the flow and the voice duration into the flow package and the voice package according to the preset voice package and the flow package, forming a basic package and recommending the basic package to a user. For example, the traffic packages are A1, A2, and A3 packages, respectively, and the corresponding traffic is 5G, 10G, and 30G, respectively; the voice packages are set B1, set B2 and set B3 respectively, and the corresponding voice durations are respectively 30 minutes, 80 minutes and 150 minutes. And when the flow of the prediction result is 9G and the voice duration is 70 minutes, combining the flow package A2 and the voice package B2 into a basic package and recommending the basic package to a corresponding user.

Illustratively, the recommendation to the user may be via short message, telephone, or APP popup. Specifically, how to recommend a package to a user can be determined by those skilled in the art according to actual situations, and the present invention is not limited herein.

In the embodiment of the invention, user behavior data is obtained from log data according to the mobile phone number of a user; screening characteristic data from the user behavior data, and performing a prediction task aiming at voice use and a prediction task aiming at flow use through a multi-task model by adopting the characteristic data to obtain a prediction result; and determining the voice package and the flow package recommended for the user according to the prediction result. By using the multi-task model, the flow and the voice tasks are predicted respectively, the prediction efficiency is improved, the package of the most suitable user can be matched more accurately, the service is provided to the greatest extent, and the user is reserved.

Referring to fig. 3, a flowchart of a training method of a multitask model in an embodiment of the present invention is shown, where the training method of the multitask model includes:

step 301, obtaining feature sample data, wherein the feature sample data comprises a basic feature sample, an interactive feature sample and a time slice feature sample;

for example, the feature sample data may include a basic feature sample, an interactive feature sample, a time slice feature sample, and the like, and the obtaining of the feature sample data may be performed by filtering the historical package use data of the user, specifically, how to obtain the feature sample data, which may be determined by a person skilled in the art according to an actual situation, and the present invention is not limited herein.

Step 302, using the feature sample data as the input of the multitask model;

step 303, in the multitask model, performing a prediction task for voice use and a prediction task for flow use by using the feature sample data to obtain a prediction result;

in the embodiment of the invention, the acquired feature sample data can be used as the input of the multitask model, so that the prediction result aiming at the voice and the flow can be obtained through the multitask model. The multi-task model is modeled based on a voice task and a flow task respectively; the corresponding prediction results can be voice duration and flow, and specifically, the Tower network can be set based on the prediction demands of different tasks, so as to obtain the prediction results corresponding to the demands.

In an alternative embodiment, the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; said step 303 may comprise the following sub-steps S31-S33:

substep S31, processing the feature sample data through each gate network to obtain the weight of a plurality of expert networks corresponding to each gate network;

in an alternative embodiment, the substep S31 may comprise:

performing feature extraction on the feature sample data through each gate network to obtain the probability of selecting a plurality of expert networks by each gate network; and carrying out normalization processing on the probability of the plurality of expert networks selected by each gate network to obtain the weight of the plurality of expert networks corresponding to each gate network.

In the embodiment of the invention, each gate network can be used for performing feature extraction on input feature sample data to obtain a second feature sample of the input feature data; performing softmax processing on the second feature sample through the gate network to obtain probabilities of the plurality of expert networks being selected by each gate network; wherein the probability may be 0 in magnitude.

The weights of the plurality of expert networks corresponding to each gate network are obtained by normalizing the probabilities of the plurality of expert networks corresponding to each gate network.

Step S32, respectively extracting the characteristics of the characteristic sample data through a plurality of expert networks corresponding to each gate network to obtain a plurality of first characteristic samples;

substep S33, according to the weights of the expert networks corresponding to each gate network, carrying out weighted summation on a plurality of first characteristic samples corresponding to each gate network, and obtaining a prediction result of flow and voice duration through the Tower network corresponding to each gate network;

in the embodiment of the invention, each gate network can correspond to a plurality of expert networks, and the expert networks are used for performing feature extraction on input feature sample data to obtain a plurality of corresponding first feature samples. Wherein the number of first feature samples is the same as the number of expert networks.

And performing weighted summation on the plurality of first characteristic samples according to the weight of the plurality of expert networks corresponding to each gate network, and obtaining the prediction results of the flow and the voice duration through the Tower network corresponding to each gate network.

And step 304, training the multitask model according to the prediction result.

In the embodiment of the invention, the multi-task model can be trained according to the corresponding prediction result; illustratively, the trained multi-task model can be screened in a ten-fold cross validation manner or a five-fold cross validation manner.

In the embodiment of the invention, in the multitask model, the voice use prediction task and the flow use prediction task are carried out through the characteristic sample data to train the multitask model, so that the model robustness can be improved, and the voice task and the flow task are simultaneously predicted through different gate networks, thereby improving the prediction accuracy and the prediction efficiency.

It should be noted that, for simplicity of description, the method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the illustrated order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments of the present invention. Further, those of skill in the art will appreciate that the embodiments described in the specification are presently preferred and that no particular act is required to implement the embodiments of the invention.

Referring to fig. 4, a block diagram of a structure of a recommendation apparatus based on a multitask model according to an embodiment of the present invention is shown, which may specifically include the following modules:

the data acquiring module 401 is configured to acquire user behavior data from log data according to a mobile phone number of the user;

a feature screening module 402, configured to screen feature data from the user behavior data, where the feature data includes a basic feature, an interactive feature, and a time slice feature;

a feature prediction module 403, configured to perform a prediction task for voice usage and a prediction task for traffic usage through a multi-task model by using the feature data to obtain a prediction result;

a package determining module 404, configured to determine, according to the prediction result, a voice package and a flow package recommended for the user.

In one embodiment, the package determination module comprises:

In one embodiment, the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; the feature prediction module includes:

the feature extraction sub-module is used for respectively extracting features of the feature data through a plurality of expert networks corresponding to each gate network to obtain a plurality of first features;

and the prediction result submodule is used for carrying out weighted summation on a plurality of first characteristics corresponding to each gate network according to the weights of a plurality of expert networks corresponding to each gate network, and obtaining the prediction result through the Tower network corresponding to each gate network.

In one embodiment, the weight obtaining sub-module includes:

In one embodiment, the multitask model is trained by:

the sample prediction module is used for performing a prediction task aiming at voice use and a prediction task aiming at flow use by adopting the characteristic sample data in the multitask model to obtain a prediction result;

In one embodiment, the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; the sample prediction module comprises:

In one embodiment, the sample weight submodule includes:

In summary, in the embodiment of the present invention, user behavior data is obtained from log data according to the mobile phone number of the user; screening characteristic data from the user behavior data, and performing a prediction task aiming at voice use and a prediction task aiming at flow use through a multi-task model by adopting the characteristic data to obtain a prediction result; and determining the voice package and the flow package recommended for the user according to the prediction result. By using the multi-task model, the flow and the voice tasks are predicted respectively, the prediction efficiency is improved, the package of the most suitable user can be matched more accurately, the service is provided to the maximum extent, and the user is kept.

For the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and reference may be made to the partial description of the method embodiment for relevant points.

An embodiment of the present invention further provides an electronic device, including:

the recommendation method based on the multi-task model comprises a processor, a memory and a computer program which is stored in the memory and can run on the processor, wherein when the computer program is executed by the processor, each process of the recommendation method based on the multi-task model is realized, the same technical effect can be achieved, and in order to avoid repetition, the description is omitted here.

The embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program implements each process of the above-mentioned recommendation method embodiment based on a multitask model, and can achieve the same technical effect, and in order to avoid repetition, the details are not repeated here.

The embodiments in the present specification are all described in a progressive manner, and each embodiment focuses on differences from other embodiments, and portions that are the same and similar between the embodiments may be referred to each other.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While preferred embodiments of the present invention have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the embodiments of the invention.

Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrases "comprising one of \ 8230; \8230;" does not exclude the presence of additional like elements in a process, method, article, or terminal device that comprises the element.

The recommendation method based on the multitask model and the recommendation device based on the multitask model provided by the invention are described in detail, specific examples are applied in the description to explain the principle and the implementation mode of the invention, and the description of the examples is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims

1. A recommendation method based on a multitask model is characterized by comprising the following steps:

2. The method of claim 1, wherein determining the recommended voice packages and traffic packages for the user based on the prediction comprises:

if the prediction result is the flow and the voice time length, mapping the flow and the voice time length into a voice package and a flow package according to a preset mapping relation;

3. The method of claim 1, wherein the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; the step of obtaining a prediction result by adopting the characteristic data and performing a prediction task for voice use and a prediction task for flow use through a multi-task model comprises the following steps:

4. The method of claim 3, wherein the processing the feature data through each gate network to obtain weights for a plurality of expert networks corresponding to each gate network comprises:

5. The method of claim 1, wherein the multitask model is trained by:

taking the feature sample data as an input of the multitask model;

and training the multitask model according to the prediction result.

6. The method of claim 5, wherein the multitasking model comprises: the system comprises two gate networks, a plurality of expert networks and two Tower networks, wherein the two gate networks correspond to the two Tower networks one by one; in the multitask model, performing a prediction task for voice use and a prediction task for traffic use by using the feature sample data to obtain a prediction result, including:

processing the feature sample data through each gate network to obtain the weight of a plurality of expert networks corresponding to each gate network;

7. The method of claim 6, wherein said processing said feature sample data through each of said portal networks to obtain weights for a plurality of said expert networks corresponding to each of said portal networks comprises:

8. A recommendation device based on a multitasking model, the device comprising:

the characteristic prediction module is used for carrying out a prediction task aiming at voice use and a prediction task aiming at flow use through a multi-task model by adopting the characteristic data to obtain a prediction result;

9. An electronic device, comprising: processor, memory and a computer program stored on the memory and being executable on the processor, the computer program, when executed by the processor, implementing the steps of the multitask model based recommendation method according to any one of claims 1-7.

10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the multitask model based recommendation method according to any one of claims 1-7.