CN114936907A - Commodity recommendation method and system based on node type interaction - Google Patents

Abstract

The invention discloses a product recommendation method and system based on node type interaction. Map all nodes to the same feature space; build type interaction functions between users-commodities, users-stores, and commodity-stores to perform different types of type interactions on the node features after feature conversion, and according to the edge type, after the type interaction The weighted nodes are weighted; the neighbor node information is aggregated for the weighted nodes, so as to update the heterogeneous information network, and the updated heterogeneous information network is used to recommend products according to the product recommendation task. Not limited to the meta-path-based recommendation method, feature transformation and interaction functions are designed for node types, so that the acquired network information is more comprehensive, and the problems of data sparseness and cold start are alleviated.

Description

A product recommendation method and system based on node type interaction

technical field

The present invention relates to the technical field of recommendation systems, in particular to a method and system for recommending products based on node type interaction.

Background technique

The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

The recommendation system learns user preferences, digs deep into the potential needs of users, predicts the items that users are interested in, and makes personalized recommendations for users. On the Internet, people can browse and purchase items that meet their own requirements from a variety of items. However, with the advent of the era of big data, the amount of information has exploded, and large volumes and various types of data are mixed together. It provides users with more purchasing options. On the other hand, it also increases the difficulty for users to select items and filter information. Therefore, the method and system for recommending items aims to analyze the different needs of each user, and make recommendations for users from many items. Personalized selection and recommendation can alleviate the problem of information overload and improve the user's shopping experience.

Commodity-based recommendation system belongs to a heterogeneous information network, which consists of various types of nodes and various types of connecting edges. For example, nodes include items, users, etc., and connecting edges include clicks, add-ons, and attention. Nodes are analyzed from different perspectives and have different attribute information. Items can have attribute information such as clothes, food, and skin care products, or attribute information such as essential and non-essential products. Users can have attribute information such as teenagers, middle-aged and elderly, or students, Attribute information such as office workers, stores can have attribute information such as men's clothing and women's clothing, or attribute information such as clothing and food.

The basic steps of product recommendation are to first model the user-product interaction data and all auxiliary information as a heterogeneous information network, and then extract the network information to design an appropriate recommendation model. Recommendations based on heterogeneous information networks usually use hand-designed meta-paths to extract network information. However, this method is highly complex, and due to the rich semantic information contained in heterogeneous information networks, it is often difficult to exhaust multiple meta-paths. In the process of finding the neighbor nodes of the target vertex, the problem of information loss is easy to occur.

SUMMARY OF THE INVENTION

In order to solve the above problems, the present invention proposes a product recommendation method and system based on node type interaction. In order to improve the comprehensiveness and accuracy of the information obtained by the recommendation system, it is not limited to the recommendation method based on meta-path, and design features for node types. The transformation and interaction functions make the obtained network information more comprehensive and alleviate the problems of data sparseness and cold start.

In order to achieve the above object, the present invention adopts the following technical solutions:

In a first aspect, the present invention provides a product recommendation method based on node type interaction, including:

Build a heterogeneous information network with commodity types, stores and user types as nodes and user behavior as edges;

After feature transformation of different types of nodes, all nodes are mapped to the same feature space;

Construct user-commodity, user-store, and commodity-store type interaction functions to perform different types of type interactions on node features after feature conversion, and weight the nodes after type interaction according to edge types;

The weighted nodes are aggregated with neighbor node information to update the heterogeneous information network, and the updated heterogeneous information network is used to recommend products according to the product recommendation task.

以使每个节点映射到d维向量；节点i经过特征转换后的特征

为：

其中，h_i为节点i的初始特征。As an alternative implementation, each type of node is designed with a feature transfer function

so that each node is mapped to a d-dimensional vector; the feature of node i after feature transformation

for:

Among them, h _i is the initial feature of node i.

As an optional implementation, the type interaction function H( _au , a _g ) between users and commodities is:

The type interaction function H(a _u , a _s ) between users and stores is:

The type interaction function H(a _g , a _s ) between commodities and stores is:

Among them, k _j and k _i respectively represent two different types of node features for type interaction.

As an optional implementation manner, the node characteristics after type interaction include:

For node i of user type, the feature h _i '(u) after type interaction is expressed as:

For node i of commodity type, the feature h _i '(g) after type interaction is expressed as:

For node i of store type, the feature h _i '(s) after type interaction is expressed as:

为节点i经过特征转换后的特征。in,

is the feature of node i after feature transformation.

其中，h_i′为类型交互后的节点特征，

是归一化后的权重。As an optional embodiment, the feature h _i " of the node after weighting is expressed as:

Among them, h _i ′ is the node feature after type interaction,

is the normalized weight.

其中，W是权重矩阵，b是偏置向量。As an optional implementation, the weights are:

where W is the weight matrix and b is the bias vector.

表示为：As an optional implementation, the features after aggregating neighbor node information

Expressed as:

为节点i经过特征转换后的特征，H(a_i,a_j)为类型a_i和类型a_j的类型交互函数。Among them, h _i ″ is the feature of the node after weighting,

is the feature of node i after feature transformation, H( _ai , a _j ) is the type interaction function of type a _i and type a _j .

In a second aspect, the present invention provides a product recommendation system based on node type interaction, including:

The network building module is configured to construct a heterogeneous information network with commodity types, stores and user types as nodes and user behavior as edges;

The feature transformation module is configured to map all nodes to the same feature space after feature transformation for different types of nodes;

The type interaction module is configured to construct user-commodity, user-store, and commodity-store type interaction functions, so as to perform different types of type interactions on the node features after feature conversion, and according to the edge type. to empower;

The information aggregation module is configured to aggregate the neighbor node information for the weighted nodes, so as to update the heterogeneous information network, and use the updated heterogeneous information network to perform product recommendation according to the product recommendation task.

In a third aspect, the present invention provides an electronic device, comprising a memory, a processor, and computer instructions stored in the memory and executed on the processor, and when the computer instructions are executed by the processor, the method described in the first aspect is completed .

In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, and when the computer instructions are executed by a processor, the method described in the first aspect is completed.

Compared with the prior art, the beneficial effects of the present invention are:

The present invention proposes a product recommendation method and system based on the interaction of node types. The type interaction function is designed according to the node type, so that the acquired network information is more comprehensive, the problems of data sparseness and cold start are alleviated, and the recommended results are also interpretable. And there are traces.

Big data contains rich semantic information, and extracting semantic information through meta-paths is generally to select meta-paths with rich connections and strong semantic characteristics, but finding such meta-paths requires more domain knowledge. The invention gets rid of the dependence on the meta-path, avoids the complexity of manually designing the meta-path, and extracts a more comprehensive network information representation.

The method and system for commodity recommendation based on node type interaction proposed by the present invention focus on extracting node information, and learn rich semantic relationships for recommendation tasks by maintaining the structural characteristics of the network mode.

Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will become apparent from the description which follows, or may be learned by practice of the invention.

Description of drawings

The accompanying drawings forming a part of the present invention are used to provide further understanding of the present invention, and the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute an improper limitation of the present invention.

1 is a flowchart of a method for recommending products based on node type interaction provided in Embodiment 1 of the present invention;

FIG. 2 is a heterogeneous information network diagram provided by Embodiment 1 of the present invention.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and embodiments.

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the invention. Unless otherwise defined, 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.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that the terms "including" and "having" and any conjugations thereof are intended to cover the non-exclusive A process, method, system, product or device comprising, for example, a series of steps or units is not necessarily limited to those steps or units expressly listed, but may include those steps or units not expressly listed or for such processes, methods, Other steps or units inherent to the product or equipment.

Embodiments of the invention and features of the embodiments may be combined with each other without conflict.

Example 1

This embodiment provides a product recommendation method based on node type interaction. In order to improve the comprehensiveness and accuracy of the information obtained by the recommendation system, the interaction function is designed for the node type without being restricted to the recommendation method based on the meta-path, so that the obtained network The information is more comprehensive, and the problems of data sparseness and cold start are alleviated.

As shown in Figure 1, it includes:

Build a heterogeneous information network with commodity types, stores and user types as nodes and user behavior as edges;

After feature transformation of different types of nodes, all nodes are mapped to the same feature space;

Construct user-commodity, user-store, and commodity-store type interaction functions to perform different types of type interactions on node features after feature conversion, and weight the nodes after type interaction according to edge types;

The weighted nodes are aggregated with neighbor node information to update the heterogeneous information network, and the updated heterogeneous information network is used to recommend products according to the product recommendation task.

In this embodiment, the commodity types include food, beauty, clothing, stationery, digital, etc., users include young men and women and children, the store corresponds to the type of commodity, and user behaviors include purchase, additional purchase, collection, listing, attention, etc. , the constructed heterogeneous information network is shown in Figure 2.

其中

表示节点集合，ε表示边集合；The heterogeneous information network is represented as

in

represents the node set, and ε represents the edge set;

其中

表示节点类型的集合，

表示连接边类型的集合；对于两个相邻节点

设其节点类型分别为a_i、a_j，连接边为e＝(i,j,r_i,j)∈ε，连接边类型为r_i,j。Let the node type mapping function and the connection edge type mapping function be respectively

in

represents a collection of node types,

Represents a collection of connected edge types; for two adjacent nodes

Let its node types be a _i , a _j respectively, the connecting edge is e=(i,j,r _i,j )∈ε, and the connecting edge type is ri _,j .

It is understandable that all the above data are obtained on the basis of compliance with laws and regulations and the user's consent, and the data is legally used.

中节点i、j的初始特征分别为h_i、h_j，首先对节点类型分类为

为每一类型的节点设计特征转换函数

使得

将所有类型节点映射到同一特征空间中，即将每个节点都映射到d维向量；In this embodiment, a feature conversion function C is designed for different types of nodes of commodity type, store and user type; because the recommended data types are complex and the feature spaces of different types of nodes are different, the graph

The initial features of nodes i and _{j are respectively hi and h j .} First, the node types are classified as

Design feature transfer functions for each type of node

make

Map all types of nodes into the same feature space, that is, map each node to a d-dimensional vector;

Specifically, the features of nodes i and j after feature transformation are expressed as:

In this embodiment, the type interaction functions H(a _u , a _g ), H(a _u , a _s ) and H(a _g , a _s ) between user-commodity, user-store, and product-store are constructed respectively. , to learn the correlation and conversion relationship between users, products and stores, respectively;

Specifically, the type interaction function H(a _u , a _g ) between user-commodity nodes is:

The type interaction function H(a _u , a _s ) between user-store nodes is:

The type interaction function H(a _g , a _s ) between commodity-store nodes is:

Among them, k _j and k _i respectively represent two different types of node features for type interaction.

For the node i belonging to the user type, its interactive features are expressed as:

For the node i belonging to the commodity type, its interactive features are expressed as:

For node i belonging to the store type, its interactive features are expressed as:

In this embodiment, the nodes after type interaction are weighted according to the edge type; namely:

是归一化后的权重；in,

is the normalized weight;

where W is the weight matrix and b is the bias vector.

In this embodiment, neighbor node information is aggregated for the weighted nodes to obtain high-order semantic information of the nodes. Neighbor nodes are nodes connected to node i, aggregate the information of nodes connected to node i, and add a self-loop to each node, so that nodes retain their own characteristics, extract richer attribute information of the network, and at the same time for cold Startup issues are also somewhat mitigated.

The features after aggregating neighbor node information are expressed as:

In this embodiment, according to the above processing of nodes, the heterogeneous information network is updated, and the cross-entropy loss function is used for optimization, so that after the products are sorted according to the product recommendation task, the recommendation is carried out in combination with the top-k recommendation strategy.

Example 2

This embodiment provides a product recommendation system based on node type interaction, including:

The network building module is configured to construct a heterogeneous information network with commodity types, stores and user types as nodes and user behavior as edges;

The feature transformation module is configured to map all nodes to the same feature space after feature transformation for different types of nodes;

The type interaction module is configured to construct user-commodity, user-store, and commodity-store type interaction functions, so as to perform different types of type interactions on the node features after feature conversion, and according to the edge type. to empower;

The information aggregation module is configured to aggregate the neighbor node information for the weighted nodes, so as to update the heterogeneous information network, and use the updated heterogeneous information network to perform product recommendation according to the product recommendation task.

It should be noted here that the foregoing modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the foregoing modules and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that the above modules may be executed in a computer system such as a set of computer-executable instructions as part of the system.

In further embodiments, there is also provided:

An electronic device includes a memory, a processor, and computer instructions stored on the memory and executed on the processor, and when the computer instructions are executed by the processor, the method described in Embodiment 1 is completed. For brevity, details are not repeated here.

It should be understood that, in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate array FPGA or other programmable logic devices , discrete gate or transistor logic devices, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory may include read-only memory and random access memory and provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

A computer-readable storage medium for storing computer instructions, when the computer instructions are executed by a processor, the method described in Embodiment 1 is completed.

The method in Embodiment 1 may be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor. The software modules may be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media mature in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware. To avoid repetition, detailed description is omitted here.

Those of ordinary skill in the art can realize that the unit, that is, the algorithm step of each example described in conjunction with this embodiment, can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

Although the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, they do not limit the scope of protection of the present invention. Those skilled in the art should understand that on the basis of the technical solutions of the present invention, those skilled in the art do not need to pay creative efforts. Various modifications or deformations that can be made are still within the protection scope of the present invention.

Claims (10)

1. A product recommendation method based on node type interaction, characterized in that, comprising: Build a heterogeneous information network with commodity types, stores and user types as nodes and user behavior as edges; After feature transformation of different types of nodes, all nodes are mapped to the same feature space; Construct user-commodity, user-store, and commodity-store type interaction functions to perform different types of type interactions on node features after feature conversion, and weight the nodes after type interaction according to edge types; The weighted nodes are aggregated with neighbor node information to update the heterogeneous information network, and the updated heterogeneous information network is used to recommend products according to the product recommendation task. 2. a kind of product recommendation method based on node type interaction as claimed in claim 1 is characterized in that, each type of node design feature conversion function

so that each node is mapped to a d-dimensional vector; the feature of node i after feature transformation

for:

Among them, h _i is the initial feature of node i. 3. A kind of product recommendation method based on node type interaction as claimed in claim 1, is characterized in that, the type interaction function H(a _u , a _g ) between user and product is:

The type interaction function H(a _u , a _s ) between users and stores is:

The type interaction function H(a _g , a _s ) between commodities and stores is:

Among them, k _j and k _i respectively represent two different types of node features for type interaction. 4. A product recommendation method based on node type interaction according to claim 1, wherein the node characteristics after type interaction include: For node i of user type, the feature h _i '(u) after type interaction is expressed as:

For node i of commodity type, the feature h _i '(g) after type interaction is expressed as:

For node i of store type, the feature h _i '(s) after type interaction is expressed as:

in,

is the feature of node i after feature transformation. 5. A product recommendation method based on node type interaction according to claim 1, characterized in that, the feature h _i ″ of the node after weighting is expressed as: h _i ″=(β _ri +1) h _i ′ ; among them, _hi ′ is the node feature after type interaction, and β _ri is the normalized weight. 6. a kind of product recommendation method based on node type interaction as claimed in claim 5 is characterized in that, weight is:

where W is the weight matrix and b is the bias vector. 7. A product recommendation method based on node type interaction as claimed in claim 1, characterized in that the feature after aggregating neighbor node information

Expressed as:

Among them, h _i ″ is the feature of the node after weighting,

is the feature of node i after feature transformation, H( _ai , a _j ) is the type interaction function of type a _i and type a _j . 8. A product recommendation system based on node type interaction, comprising: The network building module is configured to construct a heterogeneous information network with commodity types, stores and user types as nodes and user behavior as edges; The feature transformation module is configured to map all nodes to the same feature space after feature transformation for different types of nodes; The type interaction module is configured to construct user-commodity, user-store, and commodity-store type interaction functions, so as to perform different types of type interactions on the node features after feature conversion, and according to the edge type. to empower; The information aggregation module is configured to aggregate the neighbor node information for the weighted nodes, so as to update the heterogeneous information network, and use the updated heterogeneous information network to perform product recommendation according to the product recommendation task. 9. An electronic device, characterized in that, comprising a memory and a processor and a computer instruction stored on the memory and run on the processor, when the computer instruction is run by the processor, completes any one of claims 1-7 the method described. 10 . A computer-readable storage medium, characterized in that it is used for storing computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1-7 is completed. 11 .

Images