An Improved Hybrid Recommendation System Algorithm for Resolving the Cold-Start Issues

Abstract

Online shopping has turned out to be very popular nowadays. Recommendation systems are decision aids to analyze customer’s purchase sequences and their product information to provide customer preferences A sequential pattern mining method called the Prefix Span algorithm is used to find common sub-sequences that are longer than the minimal support requirements. Rules are constructed using frequent sequences to improve the performance for identify top-N prediction. A significant challenge faced by recommendation systems is the cold-start problem. The issue arises when the system does not have enough information to propose new users. This work tries to solve the issue of cold starting by incorporating sequential rules with the Bi-clustering approach. The recommendation system is evaluated using Precision, Recall, F1 measure and accuracy. Our investigation revealed that incorporating bi-clustering enhances performance and effectively resolves the cold-start problem.