CN112328646A

CN112328646A - Multitask course recommendation method and device, computer equipment and storage medium

Info

Publication number: CN112328646A
Application number: CN202110000772.5A
Authority: CN
Inventors: 杨德杰
Original assignee: Ping An Technology Shenzhen Co Ltd
Current assignee: Ping An Technology Shenzhen Co Ltd
Priority date: 2021-01-04
Filing date: 2021-01-04
Publication date: 2021-02-05
Anticipated expiration: 2041-01-04
Also published as: CN112328646B

Abstract

The invention relates to the technical field of artificial intelligence, and provides a method, a device, computer equipment and a storage medium for recommending multitask courses, wherein the method comprises the following steps: generating a target characteristic value for each user under each task; extracting a plurality of first target user features contributing to a plurality of tasks from a plurality of user features of a plurality of users and generating a first shared feature vector based on the plurality of first target user features of each user; identifying and obtaining the task weight of each task based on the task description of each task and initializing a target risk loss function according to the task weight; iteratively training a plurality of fully-connected neural networks based on a plurality of first shared feature vectors and a target risk loss function, and simplifying the plurality of fully-connected neural networks according to a preset simplification strategy in an iterative process to obtain a multi-task prediction model; course recommendations are made using a multi-tasking predictive model. The multi-task prediction model has high prediction accuracy, improves the course recommendation effect and gives consideration to the multi-task target.

Description

Multitask course recommendation method and device, computer equipment and storage medium

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a multitask course recommending method and device, computer equipment and a storage medium.

Background

One of the important goals of educational training is to implement differentiated training course recommendation according to the individual requirements of users, and the maximization of the training effect can be realized under the condition of limited training resources only if the training requirements of different individuals are met.

In the prior art, course recommendation is realized by designing a set of recommendation algorithm engine, and the quality of a recommendation algorithm is directly related to influence on the quality of a recommendation result. In the life insurance agent training scenario, there are generally two: the starting point of performance improvement is generally to aim at the ability short board of the agent, and the corresponding course is pertinently distributed in combination with the performance target, which can be called as 'letting you learn'; the starting point of the user satisfaction is that the courses are pushed in a targeted mode according to the self interest points of the agents, and the method can be called 'I want to learn'.

The inventor finds that in the process of implementing the invention, the recommendation algorithm engine in the prior art is only driven by a certain target, so that a lot of useful information is lost, single target measurement is incomplete, recommendation deviation is easily caused, the recommendation effect is poor, and differential training course recommendation aiming at the personalized requirements of users cannot be accurately implemented.

Disclosure of Invention

In view of the above, there is a need for a method, an apparatus, a computer device and a storage medium for recommending multitask courses, which can improve the accuracy of the multi-task prediction model, improve the course recommending effect and achieve the multi-task goal.

A first aspect of the present invention provides a method for multi-tasking course recommendation, the method comprising:

acquiring sample data of a plurality of users under a plurality of tasks, and generating a target characteristic value for each user under each task according to the sample data;

extracting a plurality of first target user features contributing to the tasks from the user features of the users, and generating a first shared feature vector based on the first target user features of each user;

identifying and obtaining the task weight of each task based on the task description of each task;

initializing a target risk loss function according to the task weights of the tasks;

iteratively training a plurality of fully-connected neural networks based on the plurality of first shared feature vectors and the target risk loss function, and simplifying the plurality of fully-connected neural networks according to a preset simplification strategy in an iterative process to obtain a multi-task prediction model;

and performing course recommendation by using the multi-task prediction model.

In an optional embodiment, the reducing the plurality of fully-connected neural networks according to a preset reduction strategy in the iterative process to obtain the multi-task prediction model includes:

aiming at each iterative training, obtaining the output of the current layer neuron in each fully-connected neural network;

judging whether the output of the neuron in the current layer conforms to Gaussian distribution or not;

when the output of the neurons in the current layer accords with Gaussian distribution, reducing the number of the neurons in the current layer according to a preset reduction strategy;

taking the output of the retained neuron as the input of the neuron in the next layer of the current layer;

judging whether the output of the next layer of neurons accords with Gaussian distribution;

when the output of the next layer of neurons accords with Gaussian distribution, reducing the number of the next layer of neurons according to a preset proportion;

and repeating the process until the risk loss value of the target risk loss function is calculated in an iterative mode to be minimum, stopping training the fully-connected neural networks, and obtaining the multi-task prediction model.

In an alternative embodiment, said using said multi-tasking predictive model for course recommendation comprises:

obtaining a plurality of second target user characteristics of the user to be recommended according to the plurality of first target user characteristics;

generating a second shared feature vector based on the plurality of second target user features;

predicting by using the multi-task prediction model based on the second shared characteristic vector to obtain a course prediction vector of each task;

obtaining course prediction values with the same index in the course prediction vector of each task;

adding and calculating course predicted values with the same index to obtain a course comprehensive predicted value;

and recommending courses for the user to be recommended according to the comprehensive course prediction value.

In an optional embodiment, the extracting, from a plurality of user features of the plurality of users, a plurality of first target user features that each contribute to the plurality of tasks includes:

constructing a training data set according to a plurality of user characteristics of the plurality of users under each task;

training an XGBOOST model based on each of the training data sets;

acquiring a plurality of user characteristic weights output by each XGBOOST model;

extracting a plurality of candidate user characteristic weights which are larger than a preset weight threshold value from a plurality of user characteristic weights output by each XGBOOST model;

and selecting a plurality of first target user characteristics according to the plurality of candidate user characteristic weights corresponding to each XGBOOST model.

In an optional embodiment, the identifying and obtaining the task weight of each task based on the task description of each task includes:

acquiring a training target of current training;

semantically analyzing the training target to obtain a training description;

calculating a distance between the task description and the training description for each task;

and generating a plurality of task weights according to the plurality of distances and distributing the plurality of task weights to the plurality of tasks, wherein the sum of the plurality of task weights is 1.

In an optional embodiment, the generating a plurality of task weights according to a plurality of the distances and assigning the plurality of task weights to the plurality of tasks includes:

calculating a sum of distances of a plurality of said distances;

calculating the ratio of each distance to the sum of the distances;

sorting the odds and ends and sorts the distance order;

and allocating the reverse-ordered occupation ratio to a task corresponding to the distance with the same position as the occupation ratio.

In an optional embodiment, the initializing a target risk loss function according to the task weights of the plurality of tasks includes:

defining a task predicted value of each task;

initializing a task risk loss function according to the task weight of each task and the corresponding task predicted value and target characteristic value;

and calculating the task risk loss functions of the tasks to obtain a target risk loss function.

A second aspect of the present invention provides a multitask course recommending apparatus including:

the first generation module is used for acquiring sample data of a plurality of users under a plurality of tasks and generating a target characteristic value for each user under each task according to the sample data;

the second generation module is used for extracting a plurality of first target user characteristics contributing to the tasks from the user characteristics of the users and generating a first shared characteristic vector based on the first target user characteristics of each user;

the weight identification module is used for identifying and obtaining the task weight of each task based on the task description of each task;

the function definition module is used for initializing a target risk loss function according to the task weights of the tasks;

the model training module is used for iteratively training a plurality of fully-connected neural networks based on the plurality of first shared characteristic vectors and the target risk loss function, and simplifying the plurality of fully-connected neural networks according to a preset simplification strategy in an iterative process to obtain a multi-task prediction model;

and the course recommending module is used for recommending courses by using the multi-task predicting model.

A third aspect of the invention provides a computer apparatus comprising a processor for implementing the multi-tasking course recommendation method when executing a computer program stored in a memory.

A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the multitask course recommending method.

In summary, according to the method, the apparatus, the computer device and the storage medium for recommending multitask courses, a target feature value is first generated for each user under each task, so that each considered risk loss function is initialized based on the target feature value, then, task weights of each task are identified and obtained based on task descriptions of each task, and a target risk loss function is obtained by initializing according to the task weights and the risk loss function of each task; extracting a plurality of first target user features contributing to a plurality of tasks from a plurality of user features of a plurality of users, and generating a first shared feature vector based on the plurality of first target user features of each user, wherein the first shared feature vector is used as common input of a fully-connected neural network corresponding to all the tasks, so that a multi-task target can be considered when a multi-task prediction model is trained; and finally, in the process of training the multi-task prediction model, iteratively training a plurality of fully-connected neural networks based on a plurality of first shared characteristic vectors and a target risk loss function, simplifying the plurality of fully-connected neural networks according to a preset simplification strategy in the iterative process to obtain the multi-task prediction model, and because the fully-connected neural networks are simplified in the iterative process, the number of neurons in the fully-connected neural networks is reduced, the calculated amount is reduced, the training speed of the multi-task prediction model can be increased, and the training efficiency of the multi-task prediction model is improved, so that when the multi-task prediction model is used for course recommendation, the course recommendation efficiency can be improved. The multi-task prediction model is trained based on the multi-task target, the prediction accuracy of the multi-task prediction model is high, the course recommendation effect is improved, and the multi-task target is considered.

Drawings

FIG. 1 is a flowchart of a method for recommending multitask courses according to an embodiment of the present invention.

FIG. 2 is a block diagram of a multitask course recommending device according to a second embodiment of the present invention.

Fig. 3 is a schematic structural diagram of a computer device according to a third embodiment of the present invention.

Detailed Description

In order that the above objects, features and advantages of the present invention can be more clearly understood, a detailed description of the present invention will be given below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and features of the embodiments may be combined with each other without conflict.

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 multitask course recommending method provided by the embodiment of the invention is executed by the computer equipment, and accordingly, the multitask course recommending device runs in the computer equipment.

FIG. 1 is a flowchart of a method for recommending multitask courses according to an embodiment of the present invention. The method for recommending multitask courses specifically comprises the following steps, and the sequence of the steps in the flow chart can be changed and some steps can be omitted according to different requirements.

And S11, acquiring sample data of a plurality of users under a plurality of tasks, and generating a target characteristic value for each user under each task according to the sample data.

Wherein the user may be an insurance agent engaged in training.

The plurality of tasks correspond to a plurality of service scenarios, and different tasks correspond to different service scenarios. Sample data of each user under each task can be obtained from a local database, the sample data is used for representing the target of the user under the corresponding service scene, and in order to facilitate subsequent training of the multi-task prediction model, the target characteristic value is generated to uniformly represent the target of the user under different service scenes.

The present embodiment is described by taking two tasks as an example, wherein one task is the user satisfaction, the other task is whether the performance is improved, the target characteristic value Y1 represents the user satisfaction, and the target characteristic value Y2 represents whether the performance is improved.

The target feature value Y1 may be a first user satisfaction value (e.g., 1, representing user satisfaction) or a second user satisfaction value (e.g., 0, representing user dissatisfaction). The target feature value Y2 may be a first performance enhancement value (e.g., 1, indicating that the user has performed for a period of time after completing a course) or a second performance enhancement value (e.g., 0, indicating that the user has not performed for a period of time after completing a course).

And connecting the target feature values of each user under a plurality of tasks to obtain a first target feature vector [ Y1, Y2, …, YN ], wherein N represents the number of the tasks, and the first target feature vectors of the plurality of users are connected to obtain a first target feature matrix.

S12, extracting a plurality of first target user features contributing to the plurality of tasks from the plurality of user features of the plurality of users, and generating a first shared feature vector based on the plurality of first target user features of each user.

A plurality of user characteristics of each user can be extracted from a local database, wherein the user characteristics can comprise user capability characteristics, course characteristics, learning record characteristics and the like, and the agent capability characteristics can be extracted and processed from capability requirements reflecting user attitudes, knowledge, methods and the like, such as 'morning meeting attendance', 'exhibition of E treasure industry class information reading quantity', 'insurance proposal making quantity' and the like; the course characteristics can be course attribute data of the learning course, such as 'type of course', 'source of course', 'download times of course', etc.; the learning record features may refer to learning behaviors that occur on the training platform by the user.

Because the user characteristics of the users in different tasks are not completely the same, some user characteristics may exist in one task, other user characteristics may exist in another task, and still some user characteristics may exist in all tasks at the same time, in order to balance the information of the users under multiple tasks and realize accurate course recommendation, the first target user characteristics existing in all tasks need to be extracted, and thus the first shared characteristic vector is generated based on the first target user characteristics.

constructing a training data set according to a plurality of user characteristics of a plurality of users under each task;

training an XGBOOST model based on each of the training data sets;

The XGboost (extreme Gradient boosting) is a lifting tree model, and a plurality of weak classifiers are integrated together to form a strong classifier. Before the XGBoost model is trained, a number of hyper-parameters are determined, such as learning _ rate, max _ depth, subsample (the proportion of randomly sampled samples per tree), colomple _ byte (the ratio used to control the number of columns per random sample), num _ round (the number of iterations), max _ leaf _ nodes, and so on.

Inputting the training data set into the XGB OST for iterative training, and stopping training when the iterative training times are larger than a preset time threshold value to obtain the XGB OST model. The XGB OST model not only can output a prediction result, but also can output a user feature weight corresponding to each user feature in the training data set. The user characteristic weight can represent the contribution degree of the corresponding user characteristic to the XGB OST model, the larger the user characteristic weight is, the larger the contribution degree of the corresponding user characteristic to the XGB OST model is, and the smaller the user characteristic weight is, the smaller the contribution degree of the corresponding user characteristic to the XGB OST model is.

For example, assuming that there are two tasks, a first training data set is constructed according to a plurality of user characteristics of a plurality of users under a first task and a second training data set is constructed according to a plurality of user characteristics of a plurality of users under a second task; training a first XGBOOST model based on the first training dataset and a second XGBOOST model based on the second training dataset; acquiring a plurality of first user characteristic weights output by the first XGBOOST model, thereby extracting a plurality of first candidate user characteristic weights which are larger than a preset weight threshold value from the plurality of first user characteristic weights output by the first XGBOOST model; acquiring a plurality of second user characteristic weights output by the second XGBOOST model, thereby extracting a plurality of second candidate user characteristic weights which are larger than a preset weight threshold value from the plurality of second user characteristic weights output by the second XGBOOST model; and finally, acquiring a plurality of first candidate user characteristics corresponding to a plurality of first candidate user characteristic weights corresponding to the first XGBOOST model, acquiring a plurality of second candidate user characteristics corresponding to a plurality of second candidate user characteristic weights corresponding to the second XGBOOST model, and extracting the same candidate user characteristics in the plurality of first candidate user characteristics and the plurality of second candidate user characteristics to be used as first target user characteristics.

And S13, identifying and obtaining the task weight of each task based on the task description of each task.

In a course training scenario, it is naturally desirable that all tasks have positive target results, but in real life, different tasks have different emphasis points, and therefore, in order to better fit practical applications, different task weights need to be set for different tasks.

acquiring a training target of current training;

semantically analyzing the training target to obtain a training description;

In this alternative embodiment, each training period corresponds to a training target, for example, the visual field of the user is expanded, or the performance of the user is improved, and the training target of the current training period may be obtained, and the training target may be analyzed by using a natural language algorithm to obtain a training description.

The deviation degree between the training target of the current training and the plurality of tasks can be determined according to the distance between the training description and the task description, the smaller the distance between the training description and the task description is, the more the training target of the current training is determined to be more deviated from the task corresponding to the task description, and the larger the distance between the training description and the task description is, the more the training target of the current training is determined to be more deviated from the task corresponding to the task description.

According to the deviation degree between the training target and the task, different task weights are distributed to the task, the smaller the deviation degree is, the more uniform the task weight is distributed, and the larger the deviation degree is, the more nonuniform the task weight is distributed.

calculating a sum of distances of a plurality of said distances;

calculating the ratio of each distance to the sum of the distances;

sorting the odds and ends and sorts the distance order;

The larger the proportion is, the larger the distance is, the larger the deviation degree between the training target of the current training and the task is, and the smaller weight should be assigned; smaller the duty ratio, indicating smaller distance, smaller deviation between the training target and the task for the current training, larger weight should be assigned.

For example, assuming that the inverted ratios are BN, BN-1, …, B1 and the sequentially ordered distances are D1, D2, …, DN, the ratio BN is assigned to task W1 corresponding to distance D1 and the ratio B1 is assigned to task WN corresponding to distance DN.

In the optional embodiment, task weights are distributed according to the distance between the tasks and the training targets, that is, different task weights can be generated for different tasks, and the relevance between the tasks can be reflected from the task weights, so that the multi-task prediction model can be trained better, and the prediction effect of the multi-task prediction model is improved.

S14, initializing a target risk loss function according to the task weights of the tasks.

And defining a target risk loss function of the multi-task prediction model, so that the multi-task prediction model can be conveniently trained subsequently.

defining a task predicted value of each task;

In this alternative embodiment, a Multi-gate Mixture-of-Experts framework (MMOE) may be used to implement Multi-task learning, where multiple tasks of a user have certain relevance and mutual exclusivity, for example, the interest and ability appeal of the user has certain relevance and mutual exclusivity.

If the fully shared parameters and the feature vectors limit the diversity of the targets, the training effect is not ideal, and therefore, the diversity of multiple targets can be reflected through the MMOE while the underlying sharing is ensured.

An expert network (i.e., a fully connected neural network) and a gate network for controlling the weights of the expert network may be designed for each task. Illustratively, for the first task, the weights at the three expert networks are 20%, 60%, respectively (controlled with GATEA); for the second task, the weights in the three expert networks were 70%, 10%, 20%, respectively (controlled with GATEB).

The initialized target risk loss function is:

Min：Loss=（k1*L1+K2*L2+…+KN*LN），

where Ki represents the task weight of the ith task and Li represents the task risk loss function of the ith task.

Each task is a binary problem, using a cross-entropy loss function, expressed as follows:

MinKi：loss=-Yi*logYi’-（1-Yi）*log（1-Yi’），

where Yi is a target feature value of the ith task, and Yi' is a task prediction value of the ith task.

And S15, iteratively training a plurality of fully-connected neural networks based on the plurality of first shared feature vectors and the target risk loss function, and simplifying the plurality of fully-connected neural networks according to a preset reduction strategy in an iterative process to obtain a multi-task prediction model.

And taking a plurality of first shared feature vectors as the input of each fully-connected neural network, wherein the number of the fully-connected neural networks is the same as that of the tasks, and synchronously inputting the first shared feature vectors into the fully-connected neural networks for training.

The fully-connected neural networks are connected with a risk loss layer, and the risk loss layer comprises the target risk loss function.

The preset reduction strategy may be to reduce the number of neurons according to a preset ratio, for example, to remove neurons on both sides of each layer of neurons.

When the output of a layer of neurons does not accord with Gaussian distribution, the number of the layer of neurons can be reduced without a preset reduction strategy.

And iteratively calculating the risk loss value of the target risk loss function in a gradient calculation mode, and stopping training the plurality of fully-connected neural networks when the risk loss value of the target risk loss function reaches the minimum value.

In the process of training the multi-task prediction model, the fully-connected neural networks are simplified according to a preset simplification strategy, so that the number of neurons of the fully-connected neural networks is gradually reduced, the number of the neurons is reduced, the fully-connected calculation amount can be reduced, the training efficiency of the multi-task prediction model is improved, and the course recommendation efficiency is improved.

And S16, performing course recommendation by using the multi-task prediction model.

When a course needs to be recommended for a certain user, the multi-task prediction model can be used online to recommend the course for the user.

and recommending courses for the user to be recommended according to the course prediction vector of each task.

Because the multi-task prediction model is obtained by training based on the first shared feature vectors, and the first shared feature vectors are obtained according to the first target user features, in order to meet the input consistency of the multi-task prediction model, feature fields corresponding to the first target user features need to be obtained, second target user features of a user to be recommended are obtained according to the feature fields, so that a second shared feature vector is generated based on the second target user features, the second shared feature vector is input into the multi-task prediction model for prediction, and finally, a proper course is recommended for the user to be recommended according to a result obtained by prediction.

The course prediction vector of each task comprises a plurality of course prediction values, and each course prediction value represents the preference degree of the user to be recommended to the corresponding course.

In an optional embodiment, the recommending courses for the user to be recommended according to the course prediction vector of each task includes:

For a certain course, the click rate prediction has a course prediction value, the performance prediction has a course prediction value, and the two course prediction values are weighted and summed to obtain a comprehensive prediction value of the course. The comprehensive predicted value is combined with the course predicted values of the two tasks, and can represent the comprehensive predicted result (combining interest points and performance requirements) of the user to be recommended to the course.

For example, the user ability characteristics, the course characteristics and the learning record characteristics of the user A to be recommended are input into the multi-task prediction model, and a probability vector comprising a plurality of 0 s or 1 s is output through the prediction of the multi-task learning model, wherein each element in the probability vector represents the preference degree of the user A to be recommended for the corresponding course.

Assuming that 100 courses exist in the course library, outputting 100 probability values after prediction is performed by the multi-task prediction model, sorting from top to bottom, and recommending the courses with the preset number (for example, 20) to the user A to be recommended.

The multi-task course recommendation method can be applied to intelligent education to promote the construction of intelligent cities. The invention can simultaneously give consideration to a plurality of tasks, such as 'let you learn' and 'I want to learn', realizes course recommendation, and has high recommendation efficiency and good recommendation effect; the method is applied to commerce, can realize the capacity conversion from training to production, can improve the satisfaction degree of users, and has better application prospect.

It is emphasized that the multi-tasking prediction model may be stored in a node of the blockchain in order to further ensure privacy and security of the multi-tasking prediction model.

In some embodiments, the multitask course recommending device 20 may include a plurality of functional modules comprised of computer program segments. The computer program of each program segment of the multi-tasking course recommender 20 may be stored in a memory of a computing device and executed by at least one processor to perform the functions of the multi-tasking course recommendation (described in detail with respect to FIG. 1).

In this embodiment, the multitask course recommending device 20 may be divided into a plurality of functional modules according to the functions to be executed. The functional module may include: a first generation module 201, a second generation module 202, a weight identification module 203, a function definition module 204, a model training module 205 and a course recommendation module 206. The module referred to herein is a series of computer program segments capable of being executed by at least one processor and capable of performing a fixed function and is stored in memory. In the present embodiment, the functions of the modules will be described in detail in the following embodiments.

The first generating module 201 is configured to obtain sample data of multiple users under multiple tasks, and generate a target feature value for each user under each task according to the sample data.

Wherein the user may be an insurance agent engaged in training.

The second generating module 202 is configured to extract a plurality of first target user features that all contribute to the plurality of tasks from the plurality of user features of the plurality of users, and generate a first shared feature vector based on the plurality of first target user features of each user.

In an optional embodiment, the extracting, by the second generating module 202, a plurality of first target user features contributing to the plurality of tasks from the plurality of user features of the plurality of users includes:

training an XGBOOST model based on each of the training data sets;

The weight identification module 203 is configured to identify and obtain a task weight of each task based on the task description of each task.

In an alternative embodiment, the identifying, by the weight identifying module 203, the task weight of each task based on the task description of each task includes:

acquiring a training target of current training;

semantically analyzing the training target to obtain a training description;

calculating a sum of distances of a plurality of said distances;

calculating the ratio of each distance to the sum of the distances;

sorting the odds and ends and sorts the distance order;

The function definition module 204 is configured to initialize a target risk loss function according to the task weights of the plurality of tasks.

In an alternative embodiment, the function definition module 204 initializing the target risk loss function according to the task weights of the plurality of tasks includes:

defining a task predicted value of each task;

The initialized target risk loss function is:

Min：Loss=（k1*L1+K2*L2+…+KN*LN），

MinKi：loss=-Yi*logYi’-（1-Yi）*log（1-Yi’），

The model training module 205 is configured to iteratively train a plurality of fully-connected neural networks based on the plurality of first shared feature vectors and the target risk loss function, and reduce the plurality of fully-connected neural networks according to a preset reduction strategy in an iterative process to obtain a multi-task prediction model.

In an optional embodiment, the model training module 205 reduces the plurality of fully-connected neural networks according to a preset reduction strategy in an iterative process to obtain a multi-task prediction model, including:

The course recommending module 206 is configured to recommend a course using the multi-task prediction model.

In an alternative embodiment, the course recommendation module 206 using the multi-tasking prediction model for course recommendation includes:

The multi-task course recommending device can be applied to intelligent education to promote the construction of an intelligent city. The invention can simultaneously give consideration to a plurality of tasks, such as 'let you learn' and 'I want to learn', realizes course recommendation, and has high recommendation efficiency and good recommendation effect; the method is applied to commerce, can realize the capacity conversion from training to production, can improve the satisfaction degree of users, and has better application prospect.

Fig. 3 is a schematic structural diagram of a computer device according to a third embodiment of the present invention. In the preferred embodiment of the present invention, the computer device 3 includes a memory 31, at least one processor 32, at least one communication bus 33, and a transceiver 34.

It will be appreciated by those skilled in the art that the configuration of the computer device shown in fig. 3 does not constitute a limitation of the embodiments of the present invention, and may be a bus-type configuration or a star-type configuration, and that the computer device 3 may include more or less hardware or software than those shown, or a different arrangement of components.

In some embodiments, the computer device 3 is a device capable of automatically performing numerical calculation and/or information processing according to instructions set or stored in advance, and the hardware includes but is not limited to a microprocessor, an application specific integrated circuit, a programmable gate array, a digital processor, an embedded device, and the like. The computer device 3 may also include a client device, which includes, but is not limited to, any electronic product capable of interacting with a client through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device, for example, a personal computer, a tablet computer, a smart phone, a digital camera, etc.

It should be noted that the computer device 3 is only an example, and other electronic products that are currently available or may come into existence in the future, such as electronic products that can be adapted to the present invention, should also be included in the scope of the present invention, and are included herein by reference.

In some embodiments, the memory 31 has stored therein a computer program that, when executed by the at least one processor 32, performs all or a portion of the steps of the method for multi-tasking course recommendation as described. The Memory 31 includes a Read-Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), a One-time Programmable Read-Only Memory (OTPROM), an electronically Erasable rewritable Read-Only Memory (Electrically-Erasable Programmable Read-Only Memory (EEPROM)), an optical Read-Only disk (CD-ROM) or other optical disk Memory, a magnetic disk Memory, a tape Memory, or any other medium readable by a computer capable of carrying or storing data.

Further, the computer-readable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to the use of the blockchain node, and the like.

The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.

In some embodiments, the at least one processor 32 is a Control Unit (Control Unit) of the computer device 3, connects various components of the entire computer device 3 by using various interfaces and lines, and executes various functions and processes data of the computer device 3 by running or executing programs or modules stored in the memory 31 and calling data stored in the memory 31. For example, the at least one processor 32, when executing the computer program stored in the memory, may perform all or a portion of the steps of the multi-tasking course recommendation method described in embodiments of the present invention; or to implement all or a portion of the functionality of the multi-tasking course recommender. The at least one processor 32 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or may be composed of a plurality of integrated circuits packaged with the same or different functions, including one or more Central Processing Units (CPUs), microprocessors, digital Processing chips, graphics processors, and combinations of various control chips.

In some embodiments, the at least one communication bus 33 is arranged to enable connection communication between the memory 31 and the at least one processor 32 or the like.

Although not shown, the computer device 3 may further include a power supply (such as a battery) for supplying power to each component, and preferably, the power supply may be logically connected to the at least one processor 32 through a power management device, so as to implement functions of managing charging, discharging, and power consumption through the power management device. The power supply may also include any component of one or more dc or ac power sources, recharging devices, power failure detection circuitry, power converters or inverters, power status indicators, and the like. The computer device 3 may further include various sensors, a bluetooth module, a Wi-Fi module, and the like, which are not described herein again.

The integrated unit implemented in the form of a software functional module may be stored in a computer-readable storage medium. The software functional module is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a computer device, or a network device) or a processor (processor) to execute parts of the methods according to the embodiments of the present invention.

In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is only one logical functional division, and other divisions may be realized in practice.

The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.

In addition, functional modules in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional module.

It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned. Furthermore, it is obvious that the word "comprising" does not exclude other elements or that the singular does not exclude the plural. A plurality of units or means recited in the present invention can also be implemented by one unit or means through software or hardware. The terms first, second, etc. are used to denote names, but not any particular order.

Finally, it should be noted that the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, and although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for multi-tasking course recommendation, the method comprising:

and performing course recommendation by using the multi-task prediction model.

2. The method of claim 1, wherein the refining the plurality of fully-connected neural networks according to a predetermined refinement strategy in an iterative process to obtain a multi-tasking prediction model comprises:

3. The multi-tasking course recommendation method of claim 1, wherein said using said multi-tasking predictive model for course recommendation comprises:

4. The multi-tasking course recommendation method of claim 1, wherein said extracting a plurality of first target user features from a plurality of user features of the plurality of users that contribute to the plurality of tasks comprises:

training an XGBOOST model based on each of the training data sets;

5. The multi-tasking course recommendation method of any of claims 1-4, wherein said identifying task weights for each task based on the task description for each task comprises:

acquiring a training target of current training;

semantically analyzing the training target to obtain a training description;

6. The multi-tasking course recommendation method of claim 5, wherein said generating a plurality of task weights based on a plurality of said distances and assigning said plurality of task weights to said plurality of tasks comprises:

calculating a sum of distances of a plurality of said distances;

calculating the ratio of each distance to the sum of the distances;

sorting the odds and ends and sorts the distance order;

7. The multi-tasking course recommendation method of any of claims 1-4, wherein initializing a target risk loss function based on task weights for the plurality of tasks comprises:

defining a task predicted value of each task;

8. A multitask course recommending device, said device comprising:

9. A computer device, characterized in that the computer device comprises a processor for implementing the method of multitask course recommendation according to any one of claims 1-7 when executing a computer program stored in a memory.

10. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method of multi-tasking course recommendation of any of claims 1-7.