Research and Implementation of Recommendation System Based on Neural Network
Zhang Hai, Hongyan He · 2023
The recommendation system mines the user's interest preferences based on the user's interaction history on the network to complete the user's personalized recommendation. Collaborative filtering, as a typical algorithm in personalized recommendation, is also often used in recommendation systems. But collaborative filtering still has sparse and cold start problems. In this regard, researchers usually provide recommended performance by obtaining auxiliary information such as the project's own attributes. In order to solve these problems, this paper expands the user's potential interest by using the knowledge graph to combine the user's historical interaction information. Taking the user's interaction history as a starting point, iteratively form multiple entity sets in the knowledge graph, and finally form the user's potential interest distribution by superposition, which is used to predict the click probability. Through experiments on different data sets (movies, books), the algorithm confirmed its effectiveness on these issues.