Model training and tourism product recommendation methods, devices, equipment and storage media

By constructing a model based on knowledge graphs and graph neural networks, the characteristics of routes, products, and selling points are learned, solving the problem of insufficient recommendation accuracy in online travel platforms and improving user experience and order rate.

CN116432740BActive Publication Date: 2026-03-10SHANGHAI TRAVEL INFORMATION TECH CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-03-22
Publication Date
2026-03-10

AI Technical Summary

Technical Problem

Existing online travel platforms suffer from poor user experience. Traditional collaborative filtering algorithms lack effective user behavior characteristics in scenarios involving the linkage of routes and products, resulting in insufficient recommendation accuracy and increasing the cost for customers to arrange their trips.

Method used

Using a knowledge graph and graph neural network-based approach, a first knowledge graph and a second knowledge graph are constructed to represent user intent and selling point type, respectively. By training the graph neural network, the characteristics of routes and products are learned, and the products with the highest scores or the highest rankings and their selling points are recommended using route identifiers.

Benefits of technology

It improved the accuracy of OTA system queries, enhanced the user search experience, reduced the cost of trip planning for customers, and increased the user order rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116432740B_ABST
    Figure CN116432740B_ABST
Patent Text Reader

Abstract

This invention provides a model training and tourism product recommendation method, apparatus, device, and storage medium. It can acquire a first knowledge graph based on route nodes and product nodes, where edges between route nodes and product nodes represent user intent towards product nodes; and a second knowledge graph based on product nodes and selling point nodes, where edges between product nodes and selling point nodes represent selling point types. A graph neural network is trained using the first and second knowledge graphs, learning the route's scores related to products and selling points. In the application phase, the highest-scoring or top-ranked products and their selling points are identified using route identifiers and recommended to users. Therefore, using knowledge graphs can better learn the characteristics of routes related to products and selling points, improving the effectiveness of traditional collaborative filtering algorithms, enhancing the query accuracy of OTA systems, improving the search experience for OTA platform users, and increasing user order rates.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Recommendation method for introducing item category information into graph neural network

    CN115293851A