Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Commodity recommendation model for relieving data sparsity and commodity cold start

A product recommendation, data sparse technology, applied in the field of interest mining, can solve the problems of product cold start, data sparsity, etc., to achieve the effect of enhancing expression ability, improving accuracy, and improving model performance

Pending Publication Date: 2021-07-20
XIDIAN UNIV
View PDF2 Cites 11 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a commodity recommendation model that alleviates data sparsity and commodity cold start. Based on EGES technology, multi-head self-attention network and AUGRU structure, a new recommendation model GEARec is created to solve data sparsity and commodity cold start. User product cold start problem

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Commodity recommendation model for relieving data sparsity and commodity cold start
  • Commodity recommendation model for relieving data sparsity and commodity cold start
  • Commodity recommendation model for relieving data sparsity and commodity cold start

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0067] In the product recommendation system, there are often two situations, the user has only interacted with a few products and the newly launched product has no user behavior, which are called sparsity and cold start problems respectively. For these two types of problems, the prediction performance of conventional recommendation models is not very good. Therefore, in the process of modeling product similarity through user behavior sequences in this application, EGES is used to embed and fuse various auxiliary information of products. , which alleviates the sparsity and cold start problems to a large extent.

[0068] Since the user's interest will change over a long period of time, while the interest in a short period of time is usually consistent, the sequence of product interaction time intervals of more than one hour is cut. Extract the user's historical behavior sequence, and construct a directed weighted graph G={V,E} through the obtained sequence, where the vertex V is...

Embodiment 2

[0078] In the existing technology, Google's Transformer model first proposed a multi-head attention mechanism (Multi-head Attention) and successfully applied it to machine translation, achieving remarkable results. The essence of the multi-head attention mechanism is to perform the Scaled Dot-Product Attention operation in the self-attention mechanism H times, and then obtain the output result through linear transformation. The benefit of the self-attention mechanism is that it can more effectively capture the long-distance interdependent features in the sequence, and the multi-head mechanism helps the network to capture richer feature information. Therefore, this application introduces this mechanism into the product recommendation scenario, and can better extract the dependencies between products for longer user behavior sequences.

[0079] First of all, since the self-attention module does not contain any loop or convolution structure, it is impossible to capture the order ...

Embodiment 3

[0096] In the actual e-commerce shopping scene, users’ interests usually migrate very quickly. We can effectively extract the dependencies between products through the multi-head self-attention network. The overall recommendation is based on all the user’s purchase history, not the next time. Buy recommended. Therefore, we also need to screen more influential user history behavior sequences for different candidate products.

[0097] The user behavior sequence is a time-related sequence, in which there are shallow or deep dependencies, and the recurrent neural network RNN ​​and its variants have excellent performance for time series modeling, so the present invention proposes a A GRU (GRU with Attentional Update gate, AUGRU) structure based on the attention update gate, which more specifically simulates the interest evolution path related to candidate products, and also combines the attention mechanism to screen the interest evolution path .

[0098] The method of screening i...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a commodity recommendation model for relieving data sparsity and commodity cold start, relates to the technical field of interest mining, and provides a CTR prediction model GEARec for personalized recommendation of commodities. By introducing a graph embedding technology, the problems of sparsity and cold start are solved, pre-trained Embedding is input into an upper-layer deep neural network, the convergence speed of the network can be increased, and the model performance is further improved. The GEARec model is constructed by a pre-trained Embedding layer, a multi-head self-attention network layer, an AUGRU and an MLP. After the input Embedding vector representation of the user behavior sequence passes through the multi-head self-attention network layer and the AUGRU, the calculation result and other Embedding are spliced and input into the MLP to automatically learn the nonlinear relation between the features, and finally the probability that the user purchases candidate commodities is output.

Description

technical field [0001] The invention relates to the technical field of interest mining, in particular to a product recommendation model that alleviates data sparsity and product cold start. Background technique [0002] In the actual e-commerce shopping scene, the user’s interests usually migrate very quickly, and the multi-head self-attention network can effectively extract the dependencies between products. However, the existing technology is still based on the comprehensive recommendation of all the user’s purchase history. Rather than recommendations that predict the next purchase. Therefore, we also need to screen more influential user history behavior sequences for different candidate products. [0003] In the existing product recommendation system, there are often two situations, the user has only interacted with a few products and the newly launched product has no user behavior. This situation is called sparsity and cold start problem respectively. For these two ty...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06Q30/06G06K9/62G06N3/04G06N3/08
CPCG06Q30/0631G06N3/08G06N3/045G06F18/2411
Inventor 王琨丁漩
Owner XIDIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com