Recommender System Improvement Based on User’s Social Network Graph
Recommender systems are being widely used these days for having the best and most related offers tailored to users’ needs. Several methods have been introduced in the literature on recommender systems based on machine learning and data mining techniques. In this work, I have analyzed the
performance of several well-known recommender algorithms on the 2015 Yelp academic dataset. A feature of the dataset was the friends’ graph information for each of the users. I explored the friends’ graph for four different types of users including Heavyrater user, Opinionated user,
Blacksheep user and Coldstart user. I used the data of this graph to see if I can improve recommender systems based on its information.
The results showed that such a method was useful on some of the categories introduced earlier. During this project, I have evaluated five different methods such as the traditional user-based collaborative filtering, the simple Baseline method. Based on the several performance measures including RMSE, Precision, Recall, F measure, and Accuracy, the outcome revealed promising results for the Coldstart users and Opinionated users.