Integrating Randomized Suggestions into Collaborative Filtering Applied to Movie Recommendation

Collaborative Filtering is a well-known algorithm used for recommendation systems. It predicts users' preferences using historical data, including past interactions, to recommend items they might like. The algorithm looks for similar users and uses this information for possible recommendations. But this existing algorithm still faces three problems: user-cold start, lack of diversity, and popularity bias. This paper introduces a modified version of Collaborative Filtering wherein the Linear Congruent Generator (LCG) is integrated into the algorithm. The LCG was utilized from start to finish of the process. This approach addressed the user-cold start issue by using LCG and generating a set of random users and their favorite top 10 movies. With this method, the user can select their preferred random user from the list based on their top movies and then use the selected user's data to implement Collaborative Filtering. The LCG was also used to address the issue of lack of diversity and popularity bias by generating a list of movies with the highest and least ratings. The addition of LCG to Collaborative Filtering solved the challenges that the original algorithm currently struggles with. The findings were backed up by utilizing evaluation metrics, specifically the Pearson correlation coefficient, Intra-list similarity, and Novelty.