A Crowdfunding Website Project Recommendation Method Based on Project Labels and Graph Relationships

Abstract

A crowdfunding website project recommendation method based on a project tag and a graphical relationship comprises the following steps: step 1) tree-shaped tag based recommendation: dividing attributes of a crowdfunding website project, modeling the crowdfunding website project, describing a project tag by using a tree, abstracting correlations between user behavior and the project tag into a private tag tree of user, and storing a tree-shaped tag into a document-type database CouchDB; step 2) graphical relationship based recommendation: storing graphical relationship into graph database Neo4j; and step 3) performing weighting summation averaging on two sets of projects in the step 1) and the step 2), and finally obtaining k2 recommended project comprehensive results and asynchronously transmitting the results to the user. The crowdfunding website project recommendation method based on the project tag and the graphical relationship provided by the present invention solves the problems of low recommendation efficiency and low precision and the like, which a conventional personal recommendation method is faced with when dealing with a relationship between project attribute, user behavior as well as user and project in the crowdfunding platform.

Description

technical field [0001] The present invention relates to the field of machine learning, graph database and recommendation algorithm, and specifically refers to the realization of a personalized recommendation method based on item labels, user behaviors and user similarity in the crowdfunding website environment, according to the project attributes and User behavior characteristics, use the document database CouchDB to store item tags; use Neo4j database to store user relationship and project user mapping relationship, calculate user similarity according to the graph structure, and obtain recommendation results based on user similarity, and then provide fast and efficient personality Chemical project recommendation service. Background technique [0002] The original purpose of the crowdfunding website is to help people with dreams and ideas to raise desired resources (funding crowdfunding, item crowdfunding, human crowdfunding, etc.), and project initiators use text, pictures,...

AI Technical Summary

Problems solved by technology

[0013] However, in crowdfunding websites, projects often directly or indirectly belong to multiple tags, and traditional recommendation methods often ignore the constraint relationship between these tags; There are multiple connections between projects (browsing, funding, collection, etc.), and these nodes and connections form a relatively complex graph structure, which is quite different from other types of websites

It is often difficult to define these connection relationships using traditional relational databases, and cannot effectively store them, nor can they be quickly accessed and updated, which affects the user's recommendation experience

In addition, in the crowdfunding website platform scenario, many existing recommendation algorithms often fail to consider factors such as user behavior, item category, and user similarity, which affects the accuracy of user recommendation

Images