CN110837596A

CN110837596A - Intelligent recommendation method and device, computer equipment and storage medium

Info

Publication number: CN110837596A
Application number: CN201910872210.2A
Authority: CN
Inventors: 陈楚
Original assignee: Ping An Life Insurance Company of China Ltd
Current assignee: Ping An Life Insurance Company of China Ltd
Priority date: 2019-09-16
Filing date: 2019-09-16
Publication date: 2020-02-25
Anticipated expiration: 2039-09-16
Also published as: CN110837596B

Abstract

The embodiment of the application belongs to the technical field of intelligent recommendation of big data, and relates to an intelligent recommendation method, which comprises the steps of obtaining at least one characteristic value and an original characterization vector of a project, wherein the project comprises a plurality of click sequence projects and exposure projects to be learned; splicing the at least one characteristic value and the original characterization vector to obtain a new characterization vector of each click sequence project and the exposure project; combining the new characterization vectors of the each click sequence item and the exposure item together, and performing convolution depth learning to obtain convolution characteristics; splicing the convolution characteristics and the processing result of the depth network of the original characterization vector, and inputting the spliced convolution characteristics and the processing result into an output layer of a recommendation model to obtain the click probability of an exposure item; and determining whether to recommend the exposure item according to the click probability. The application also provides an intelligent recommendation device, computer equipment and a storage medium. The recommendation method and the recommendation device can improve the recommendation accuracy of the recommendation model.

Description

Intelligent recommendation method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of intelligent big data recommendation technologies, and in particular, to an intelligent recommendation method and apparatus, a computer device, and a storage medium.

Background

With the development of network technology, personalized recommendation has become one of indispensable network services in people's network life, and has also become the key point of future development of internet products. The prior art typically uses deep learning network models for personalized product recommendations, such as: the deep learning network model is used by the e-commerce during product recommendation, a wide & deep recommendation model is generally adopted, an deep network part inputs embedding (characterization vectors) of item items, the embedding (characterization vectors) of item items comprises splicing of a click sequence and the characterization vectors of exposed items and then accessing to an MLP (multi-layer sensor), the wide network inputs features except the characterization vectors of the item items, such as user gender, commodity price, commodity purchase rate, scene ID and the like, however, the network structure does not consider the implicit features between the click sequence and the exposed items, and the deep network does not consider other attribute features of the items during training, so that the phenomenon that the recommendation result is not accurate enough often occurs, and the recommendation result is inconsistent with the actual situation.

Disclosure of Invention

An embodiment of the application aims to provide an intelligent recommendation method, an intelligent recommendation device, computer equipment and a storage medium, so as to solve the problem that in the prior art, a recommendation result by using a deep learning network is inconsistent with an actual situation.

In order to solve the above technical problem, an embodiment of the present application provides an intelligent recommendation method, which adopts the following technical solutions:

acquiring at least one characteristic value and an original characterization vector of a project, wherein the project comprises a plurality of click sequence projects and exposure projects to be learned;

splicing the at least one characteristic value and the original characterization vector to obtain a new characterization vector of each click sequence project and the exposure project;

combining the new characterization vectors of the various click sequence items and the exposure items together, and performing convolution deep learning to obtain convolution characteristics;

inputting the original characterization vectors of the items into an input layer of a deep network of a recommendation model to obtain a processing result of the deep network;

splicing the convolution characteristics and the processing result of the depth network of the original characterization vector, and inputting the result into an output layer of a recommendation model to obtain the click probability of the exposure item;

and determining whether to recommend the exposure item according to the click probability.

Further, the step of splicing the at least one feature value and the original characterization vector to obtain a new characterization vector of each click sequence item and each exposure item includes:

and writing the at least one characteristic value into the original characterization vector to form a new characterization vector.

Further, the step of combining the new characterization vectors of the click sequence item and the exposure item, and performing convolution deep learning to obtain convolution features includes:

combining the new characterization vectors of each click sequence item and each exposure item together to form a convolution image;

and performing convolution calculation on the convolution image according to a preset convolution kernel to obtain convolution characteristics.

Further, the step of performing convolution calculation on the convolution image according to a preset convolution kernel to obtain convolution characteristics includes:

and writing a two-dimensional vector consisting of each click sequence item and the new characterization vector of each exposure item into an input layer of the convolutional neural network so as to enable the convolutional neural network to carry out convolution calculation and processing, and outputting the convolution characteristics.

Further, the step of inputting the original characterization vector of the item into an input layer of a deep network of a recommendation model to obtain a deep network processing result includes:

and inputting the click sequence items and the original characterization vectors of the exposure items into a depth network together for calculation to obtain the dimension reduction characterization vectors.

Further, the step of obtaining the click probability of the exposure item by splicing the convolution feature and the processing result of the deep network and inputting the result into an output layer of the recommendation model includes:

splicing the dimensionality reduction characterization vector and the convolution characteristic to form a combined characterization vector;

and inputting the combined characterization vector into an output layer of the recommendation model, and calculating by using a sigmoid function to obtain the click probability of the exposure item.

Further, the step of obtaining at least one feature value and an original token vector of the item to be learned includes:

extracting at least one feature value from the feature data according to the project;

and acquiring user data of the project, and converting the user data into a characterization vector.

In order to solve the above technical problem, an embodiment of the present application further provides an intelligent recommendation device, where the intelligent recommendation device includes:

the system comprises a vector acquisition module, a learning module and a learning module, wherein the vector acquisition module is used for acquiring at least one characteristic value and an original characterization vector of a project, and the project comprises a plurality of click sequence projects and exposure projects to be learned;

the splicing module is used for splicing the at least one characteristic value and the original characterization vector to obtain new characterization vectors of each click sequence project and the exposure project;

the deep network processing result acquisition module is used for inputting the original characterization vectors of the items into an input layer of a deep network of a recommendation model to obtain a processing result of the deep network;

the convolution characteristic acquisition module is used for combining the new characterization vectors of the various click sequence items and the exposure items together and carrying out convolution deep learning to acquire convolution characteristics;

the click rate obtaining module is used for splicing the convolution characteristics and the processing result of the depth network of the original characterization vector and then inputting the result into an output layer of a recommendation model to obtain the click probability of the exposure item;

and the recommending module is used for determining whether to recommend the exposure item according to the click probability.

In order to solve the above technical problem, an embodiment of the present application further provides a computer device, which includes a memory and a processor, where the memory stores a computer program, and the processor implements the steps of the intelligent recommendation method when executing the computer program.

In order to solve the above technical problem, an embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solutions:

the computer readable storage medium has stored thereon a computer program which, when being executed by a processor, carries out the steps of the intelligent recommendation method as described above.

Compared with the prior art, the embodiment of the application mainly has the following beneficial effects:

the new embedding vector is generated by adding the characteristic value of the item characteristic of the item (including the click sequence item and the exposure item) into the original characterization vector (embedding vector), so that the new embedding vector can reflect richer and more accurate attributes of the item, and the effect of the model is promoted. Meanwhile, the click sequence item and the exposure item form a convolution graph and are subjected to convolution deep learning, implicit characteristics between the click sequence item and the exposure item can be learned and added to the last layer of the recommendation model to calculate the click probability of the exposure item, so that the most direct influence is exerted on the model effect, and the recommendation accuracy of the recommendation model is improved.

Drawings

In order to more clearly illustrate the solution of the present application, the drawings needed for describing the embodiments of the present application will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and that other drawings can be obtained by those skilled in the art without inventive effort.

FIG. 1 is an exemplary system architecture diagram in which the present application may be applied;

FIG. 2 is a flow diagram of one embodiment of an intelligent recommendation method according to the present application;

FIG. 3 is a schematic diagram of a convolution image formed by combining the click sequence entries and the new characterization vectors for each exposure entry;

FIG. 4 is a schematic illustration of applying computed convolution features to a recommendation model;

FIG. 5 is a schematic block diagram of one embodiment of an intelligent recommendation device according to the present application;

FIG. 6 is a schematic block diagram of one embodiment of a computer device according to the present application.

Detailed Description

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 application belongs; the terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "including" and "having," and any variations thereof, in the description and claims of this application and the description of the above figures are intended to cover non-exclusive inclusions. The terms "first," "second," and the like in the description and claims of this application or in the above-described drawings are used for distinguishing between different objects and not for describing a particular order.

Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.

In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

As shown in fig. 1, the system architecture 100 may include

terminal devices

101, 102, 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the

terminal devices

101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.

The user may use the

terminal devices

101, 102, 103 to interact with the server 105 via the network 104 to receive or send messages or the like. The

terminal devices

101, 102, 103 may have various communication client applications installed thereon, such as a web browser application, a shopping application, a search application, an instant messaging tool, a mailbox client, social platform software, and the like.

The

terminal devices

101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture experts Group Audio Layer III, mpeg compression standard Audio Layer 3), MP4 players (Moving Picture experts Group Audio Layer IV, mpeg compression standard Audio Layer 4), laptop portable computers, desktop computers, and the like.

The server 105 may be a server providing various services, such as a background server providing support for pages displayed on the

terminal devices

101, 102, 103.

It should be noted that the intelligent recommendation method provided in the embodiments of the present application is generally executed by a server, and accordingly, the intelligent recommendation apparatus is generally disposed in the server.

It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

With continuing reference to FIG. 2, a flow diagram of one embodiment of a method of click probability calculation in accordance with the present application is shown. The intelligent recommendation method comprises the following steps:

step 201, at least one characteristic value and an original characterization vector of an item are obtained, wherein the item comprises a plurality of click sequence items and an exposure item.

In this embodiment, the exposure item may refer to an item exposed on a web page, and the item may include a commodity, an article, an advertisement, and the like. The exposure item is also an item for predicting the click probability, and may be an item that has been exposed or an item waiting for exposure. Different exposure items can be adopted according to different scenes, for example, an exposed item is selected in a model training stage, and when the model is used for recommendation, the item is an item waiting for the exposure to be recommended.

The click sequence items are primarily exposed items extracted from the log, which may include only items clicked or browsed by the user.

In this embodiment, the electronic device (for example, the server shown in fig. 1) on which the intelligent recommendation method operates may be connected to the terminal device through a wired connection manner or a wireless connection manner. It should be noted that the wireless connection means may include, but is not limited to, a 3G/4G connection, a WiFi connection, a bluetooth connection, a WiMAX connection, a Zigbee connection, a uwb (ultra wideband) connection, and other wireless connection means now known or developed in the future.

In practical application, a webpage (or an APP module) records item feature data of page exposure and item behavior data of a user, and the data can be stored in a designated server in a log form. When needed, corresponding data can be extracted from the server.

In this embodiment, step 201 includes:

acquiring project characteristic data and user data from a webpage log;

extracting at least one feature value from the item feature data;

converting the user data into a characterization vector.

In the present embodiment, the item feature refers to a feature related to an item. Taking a commodity as an example, the item features may include features such as price, category, purchase rate, scene id, etc., which are represented in a numerical manner (i.e., feature value), for example, the feature value is set to 1 when the category is car insurance, and the feature value is set to 2 when the category is production insurance.

User data, refers to features associated with the user, such as user ID, user location, user search keywords, purchase records, and the like.

In practical application, data of a user related to a project can be arranged from left to right according to time sequence to obtain a serialized data structure, and then the serialized data is converted into an embedding vector through an embedding technology.

And step 202, splicing the at least one characteristic value and the original characterization vector to obtain a new characterization vector of each click sequence project and the exposure project.

In this embodiment, after writing the at least one feature value into the original embedding vector, a new embedding vector may be constructed. Namely, the new embedding vector is composed of the original embedding vector + at least one eigenvalue.

In the prior art, an embedding vector is only a vector of item, and does not contain characteristics such as price, type, purchase rate and the like. The splicing mode of the step is that the feature value of the item is placed behind the item embedding vector, if the dimension of the embedding vector is 32-dimensional, the price, the category, the purchase rate and the scene id are written in, and the new embedding of the item is changed into 36-dimensional. Each feature serves as a dimension element of the new embedding vector. The embedding vector after splicing other characteristics can reflect richer and more accurate attributes of item, and is favorable for improving the effect of the model.

And 203, combining the new characterization vectors of the individual click sequence items and the exposure items, and performing convolution deep learning to obtain convolution characteristics.

In this embodiment, the step includes:

As shown in fig. 3, 3 click sequence items (click sequence item1, click sequence item2, click sequence item3) are selected, and the exposure item constitutes a convolution image, and the convolution kernel size (size) is set to 4 × 3.

Wherein each click sequence item contains 7 vectors, and the 7 vectors are composed of the original embedding vector and the feature value of the click sequence item. The exposure item also contains 7 vectors, which are composed of the original embedding vector and the feature value of the exposure item. The exposure item is stitched directly over 3 items. As can be seen from fig. 3, this is actually a two-dimensional vector of 4 x 7, where the dashed line indicates that the convolution kernel size is set to 4 x 3.

In practical applications, a Convolutional Neural Network (CNN) may be used for convolution calculation.

The CNN is composed of input and output layers and a plurality of hidden layers, which can be divided into convolutional layers, pooling layers, and fully-connected layers, wherein the convolutional layers and the pooling layers can be cycled any number of times.

Specifically, the convolutional layer is the core of the CNN, the parameters of the layer are determined by a set of learnable filters (filters), the design of the filters is determined by convolutional kernels, the part of the filter covered on the original image matrix is convolved with the convolutional kernels each time the filter moves, and then the filter is moved according to the step length until the whole matrix is covered. Briefly, convolutional layers are used to convolve the input layers to extract features at higher levels.

The pooling layer, also known as downsampling, functions to reduce data throughput while retaining useful information. The specific operation is to divide the convolutional layers into small matrixes, and each division is replaced by the largest element to obtain a smaller new matrix.

The fully connected layer is a conventional neural network and has the function of fully connecting high-level features obtained by passing through the convolutional layer and the pooling layer for multiple times (the fully connected layer is the property of the conventional neural network), and the final predicted value, namely the convolutional feature, is calculated.

An activation function layer can be arranged between the output layer and the full communication layer, and a target CNN algorithm is added with a nonlinear factor, so that the expression capability of the neural network on the model is improved.

The step of performing convolution calculation by the convolution neural network comprises the following steps:

the input layer receives a two-dimensional vector of the sequence of clicks item and the exposure item (as shown in FIG. 3);

the convolution layer performs convolution on the input layer according to a preset convolution core, and high-level features are extracted;

the pooling layer divides the high-level features obtained by the convolution layer to obtain a new matrix;

the full connection layer performs full connection on the new matrix to obtain convolution characteristics;

the output layer outputs the convolution characteristic.

And 204, inputting the original characterization vectors of the items into an input layer of a deep network of a recommendation model to obtain a deep network processing result.

In practical application, the recommendation model is a model for calculating the click probability of an exposure item, for example, a GRU (gated current Unit) deep neural network model is a common recommendation model, and the click probability of the exposure item can be calculated by training using a historical click sequence item and related data (such as item features, embedding vectors and the like) of the exposure item, and then the exposure item is sorted according to the calculated click probability, and one or more items to be exposed arranged in front are selected for exposure and displayed on a page.

In practical application, the deep neural network model includes a deep network, the deep network is a feedforward neural network, as shown in fig. 4, a left dotted line portion is the deep network, the input of the deep network includes the original embedding vector of the click sequence item and the exposure item, and after the original embedding vector is processed by the RELU (Rectified Linear Unit, RELU, modified Linear Unit) of the deep network, the relevant features of the click sequence item and the exposure item can be better mined.

In this embodiment, step 204 includes:

and inputting the click sequence items and the original characterization vectors of the exposure items into a depth network of a recommendation model together for calculation to obtain a dimension reduction characterization vector.

Step 205, splicing the convolution characteristics and the processing result of the depth network of the original characterization vector, and inputting the result into an output layer of a recommendation model to obtain the click probability (probabilies) of the exposure item.

In this embodiment, step 205 includes:

and splicing the dimensionality reduction characterization vector and the convolution characteristic and inputting the spliced dimensionality reduction characterization vector and the convolution characteristic into an output layer of a recommendation model for calculation to obtain the click probability of the exposure item.

In practical applications, the sigmoid function is also called Logistic function, and is a sigmoid function, which can map a vector into an interval of (0, 1).

Step 206, determining whether to recommend the exposure item according to the click probability.

In practical application, the click probability of each exposure item can be calculated, then the click probability of each exposure item is sequenced, and the exposure item to be recommended is determined according to the sequencing result; or setting a preset threshold value, and determining the exposure item with the click probability exceeding the threshold value as a recommended item.

According to the intelligent recommendation method, the feature values of the item features of the items (including the click sequence item and the exposure item) are added into the original embedding vector to generate a new embedding vector, so that the embedding vector can reflect richer and more accurate attributes of the items, and the effect of the model is promoted. Meanwhile, the click sequence item and the exposure item form a convolution graph and are subjected to convolution deep learning, implicit characteristics between the click sequence item and the exposure item can be learned and added to the last layer of the recommendation model to calculate the click probability of the exposure item, so that the most direct influence is exerted on the model effect, and the recommendation accuracy of the recommendation model is improved.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the computer program is executed. The storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), or a Random Access Memory (RAM).

It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and may be performed in other orders unless explicitly stated herein. Moreover, at least a portion of the steps in the flow chart of the figure may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

With further reference to fig. 5, as an implementation of the method shown in fig. 2, the present application provides an embodiment of an intelligent recommendation apparatus, where the embodiment of the apparatus corresponds to the embodiment of the method shown in fig. 2, and the apparatus may be specifically applied to various electronic devices.

As shown in fig. 5, the intelligent recommendation device 500 according to this embodiment includes: the system comprises a characteristic value acquisition module 501, a vector acquisition module 502, a splicing module 503, a convolution characteristic acquisition module 504, a deep network processing result acquisition module 505, a click rate acquisition module 506 and a recommendation module 507. Wherein:

a feature value obtaining module 501, configured to obtain at least one feature value of a project, where the project includes multiple click sequence projects and an exposure project to be learned;

a vector obtaining module 502, configured to obtain an original characterization vector of the project;

a splicing module 503, configured to splice the at least one feature value and the original characterization vector to obtain new characterization vectors of each click sequence project and the exposure project;

a convolution feature obtaining module 504, which combines the new characterization vectors of the individual click sequence items and the exposure items together, and performs convolution deep learning to obtain convolution features;

a deep network processing result obtaining module 505, configured to input the original characterization vector of the item into an input layer of a deep network of the recommendation model to obtain a processing result of the deep network;

the click rate obtaining module 506 splices the convolution characteristics and the processing result of the depth network of the original characterization vector and inputs the spliced result into an output layer of a recommendation model to obtain the click probability of the exposure item;

and a recommending module 507, configured to determine whether to recommend the exposure item according to the click probability.

In practical application, the deep neural network model includes a deep network, the deep network is a feedforward neural network, as shown in fig. 4, a left dotted line portion is the deep network, the input of the deep network includes an original embedding vector of the click sequence item and the exposure item, and after the original embedding vector is processed by a RELU (Rectified linear unit, RELU, modified linear unit) of the deep network, the relevant features of the click sequence item and the exposure item can be better mined.

In some optional implementations of this embodiment, the feature value obtaining module 501 includes:

the data acquisition submodule is used for acquiring project characteristic data and user data from the webpage log;

a feature value extraction submodule for extracting at least one feature value from the project feature data;

and the conversion submodule is used for converting the user data into a characterization vector.

Specifically, the item feature refers to a feature related to the item. Taking a commodity as an example, the item features may include features such as price, category, purchase rate, scene id, etc., which are represented in a numerical manner (i.e., feature value), for example, the feature value is set to 1 when the category is car insurance, and the feature value is set to 2 when the category is production insurance.

In some optional implementations of this embodiment, the vector obtaining module 502 is further configured to obtain user data of the project, and convert the user data into a characterization vector.

Specifically, the user data refers to features related to the user, such as a user ID, a user location, a user search keyword, a purchase record, and the like.

In some optional implementations of this embodiment, the stitching module 503 is further configured to write the at least one feature value into the original token vector to form a new token vector.

In some optional implementations of this embodiment, the convolution feature obtaining module 504 includes:

the convolution image acquisition submodule is used for combining the new characterization vectors of each click sequence item and each exposure item together to form a convolution image;

and the calculation submodule is used for carrying out convolution calculation on the convolution image according to a preset convolution kernel to obtain convolution characteristics.

Specifically, the convolution feature calculation sub-module is further configured to write a two-dimensional vector formed by each click sequence item and the new characterization vector of each exposure item into an input layer of a convolution neural network, so that the convolution neural network performs convolution calculation and processing, and outputs the convolution feature.

As shown in fig. 3, a 3-click sequence item is selected, and the exposure item constitutes a convolution image, with the convolution kernel size (size) set to 4 × 3.

In some optional implementation manners of this embodiment, the processing result obtaining module 505 of the depth network is further configured to input the respective click sequence items and the original characterization vectors of the exposure items into the depth network to calculate to obtain the dimension reduction characterization vector.

The click rate obtaining module 506 is further configured to splice the dimensionality reduction characterization vector and the convolution feature to form a combined characterization vector;

In some optional implementation manners of this embodiment, the exposure items to be recommended may be determined according to a sorting result by calculating the click probability of each exposure item, then sorting the click probability of each exposure item by the recommending module 507; a preset threshold may also be set, and the recommendation module 507 directly determines the exposure item with the click probability exceeding the threshold as the recommendation item.

The intelligent recommendation device of the embodiment generates a new embedding vector by adding the characteristic value of the item characteristic of the item (including the click sequence item and the exposure item) into the original embedding vector, so that the embedding vector can reflect richer and more accurate attributes of the item, and the promotion of the effect of the model is facilitated. Meanwhile, the click sequence item and the exposure item form a convolution graph and are subjected to convolution deep learning, implicit characteristics between the click sequence item and the exposure item can be learned and added to the last layer of the recommendation model to calculate the click probability of the exposure item, so that the most direct influence is exerted on the model effect, and the recommendation accuracy of the recommendation model is improved.

In order to solve the technical problem, an embodiment of the present application further provides a computer device. Referring to fig. 6, fig. 6 is a block diagram of a basic structure of a computer device according to the present embodiment.

The computer device 6 comprises a memory 61, a processor 62, a network interface 63 communicatively connected to each other via a system bus. It is noted that only a computer device 6 having components 61-63 is shown, but it is understood that not all of the shown components are required to be implemented, and that more or fewer components may be implemented instead. As will be understood by those skilled in the art, the computer device is a device capable of automatically performing numerical calculation and/or information processing according to a preset or stored instruction, and the hardware includes, but is not limited to, a microprocessor, an Application Specific Integrated Circuit (ASIC), a Programmable gate array (FPGA), a Digital Signal Processor (DSP), an embedded device, and the like.

The computer device can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The computer equipment can carry out man-machine interaction with a user through a keyboard, a mouse, a remote controller, a touch panel or voice control equipment and the like.

The memory 61 includes at least one type of readable storage medium including a flash memory, a hard disk, a multimedia card, a card type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 61 may be an internal storage unit of the computer device 6, such as a hard disk or a memory of the computer device 6. In other embodiments, the memory 61 may also be an external storage device of the computer device 6, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a flash Card (FlashCard), and the like, which are provided on the computer device 6. Of course, the memory 61 may also comprise both an internal storage unit of the computer device 6 and an external storage device thereof. In this embodiment, the memory 61 is generally used for storing an operating system installed in the computer device 6 and various application software, such as program codes of the intelligent recommendation method. Further, the memory 61 may also be used to temporarily store various types of data that have been output or are to be output.

The processor 62 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data Processing chip in some embodiments. The processor 62 is typically used to control the overall operation of the computer device 6. In this embodiment, the processor 62 is configured to execute the program code stored in the memory 61 or process data, for example, execute the program code of the intelligent recommendation method.

The network interface 63 may comprise a wireless network interface or a wired network interface, and the network interface 63 is typically used for establishing a communication connection between the computer device 6 and other electronic devices.

The present application further provides another embodiment, which is to provide a computer-readable storage medium storing a click probability calculation program, where the click probability calculation program is executable by at least one processor to cause the at least one processor to execute the steps of the intelligent recommendation method as described above.

Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present application.

It is to be understood that the above-described embodiments are merely illustrative of some, but not restrictive, of the broad invention, and that the appended drawings illustrate preferred embodiments of the invention and do not limit the scope of the invention. This application is capable of embodiments in many different forms and is provided for the purpose of enabling a thorough understanding of the disclosure of the application. Although the present application has been described in detail with reference to the foregoing embodiments, it will be apparent to one skilled in the art that the present application may be practiced without modification or with equivalents of some of the features described in the foregoing embodiments. All equivalent structures made by using the contents of the specification and the drawings of the present application are directly or indirectly applied to other related technical fields and are within the protection scope of the present application.

Claims

1. An intelligent recommendation method is characterized by comprising the following steps:

acquiring at least one characteristic value and an original characterization vector of a project, wherein the project comprises a plurality of click sequence projects and exposure projects;

splicing the convolution characteristics and the processing result of the deep network, and inputting the result into an output layer of the recommendation model to obtain the click probability of the exposure item;

2. The intelligent recommendation method according to claim 1, wherein the step of concatenating the at least one eigenvalue and the original token vector to obtain a new token vector for each click sequence item and each exposure item comprises:

3. The intelligent recommendation method according to claim 1, wherein the step of combining the new characterization vectors of the click sequence item and the exposure item and performing convolution deep learning to obtain convolution features comprises:

4. The intelligent recommendation method according to claim 3, wherein the step of performing convolution calculation on the convolution image according to a preset convolution kernel to obtain the convolution characteristics comprises:

5. The intelligent recommendation method according to claim 1, wherein the step of inputting the original characterization vectors of the items into an input layer of a deep network of a recommendation model to obtain a processing result of the deep network comprises:

and inputting the click sequence items and the original characterization vectors of the exposure items into the depth network together to calculate to obtain a dimensionality reduction characterization vector as a processing result of the depth network.

6. The intelligent recommendation method according to claim 5, wherein the step of obtaining the click probability of the exposure item by inputting the convolution feature and the processing result of the deep network after being spliced into an output layer of the recommendation model comprises:

7. The intelligent recommendation method according to any one of claims 1-6, wherein the step of obtaining at least one eigenvalue and original token vector of an item comprises:

acquiring project characteristic data and user data from a webpage log;

extracting at least one feature value from the item feature data;

converting the user data into an original characterization vector.

8. A recommendation device, comprising:

the system comprises a characteristic value acquisition module, a learning module and a learning module, wherein the characteristic value acquisition module is used for acquiring at least one characteristic value of a project, and the project comprises a plurality of click sequence projects and exposure projects to be learned;

the vector acquisition module is used for acquiring an original characterization vector of the project;

9. A computer device comprising a memory having stored therein a computer program and a processor implementing the steps of the intelligent recommendation method of any of claims 1-7 when executing the computer program.

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