A CROSS_DOMAIN RECOMMENDATION SYSTEM

Abstract

In today's digital era, Recommendation Systems play an important role in providing suitable recommendations to users based on their preferences and behaviors. However, traditional recommendation systems face some important problems, especially the cold-start problem and lack of diversity in the recommendation results. This project studies and proposes solutions to these problems by developing a multi-domain recommendation system, using modern methods to improve the accuracyanddiversity of the results. To solve the above problems, I have applied three main techniques: PearsonCorrelation, Cosine Similarity and K-Nearest Neighbors (KNN). These methods are combined in the recommendation system to improve the accuracy and diversity of the recommendations. Among them, Pearson Correlation is used to measure the correlation between objects, Cosine Similarity assists in calculating the similarity between targets, and KNN helps improve the prediction results, especially in situations with little initial data. The experimental results show that the combination of the above methods has significantly improved the system evaluation indexes such as Mean Absolute Error(MAE) and the accuracy of the recommendations. This proves that the multi-domain recommender system can effectively solve the cold-start and lack of diversity problems, and improve the quality of recommendations for users.