Evolutionary User Clustering Based on Time-Aware Interest Changes in the Recommender System

The plenty of data on the Internet has created problems for users and has caused confusion in finding the proper information. Also, users' tastes and preferences change over time. Recommender systems can help users find useful information. Due to changing interests, systems must be able to evolve. In order to solve this problem, users are clustered that determine the most desirable users, it pays attention to the user's rating of the items. The time parameter has been considered in the proposed method of Genetic Algorithm-Simulated Algorithm (SAGA) of this paper which can improve user prioritization based on time. In the proposed method, using the Memetic evolutionary algorithm, the clusters are improved over time, which it provides appropriate suggestions to the user. The system also performs optimal evolutionary clustering using item properties for the cold start item problem, and user demographic information for the cold start user problem. The proposed method has been evaluated using the Movielens dataset and experimental results show that the proposed SAGA method with an accuracy of 0.89 has a better performance in the accuracy of predictions and suggestions to users than existing methods.