Improving the Data Quality of the MovieLens Dataset Using Dimensionality Reduction Techniques

Abstract

The presence of recommendation systems within the current era is not controlled or even managed like before, for providing the optimum services to users. For that reason, these services need to be in an ideal and complete state since the real-world data is often not that much perfect. That optimum state could not be achieved without the interaction of preprocessing method. Preprocessing can offer a lot of insights about the data and its structure and quality. Preprocessing techniques ensure that the dataset has complete, consistent and integrant properties for further analysis. Dimensionality reduction is this paper is discussed as one of the preprocessing stages using the most popular matrix factorization methods. We have also made an experiment to show how the dimensionality reduction algorithms can overcome any constrained computational resources and prove better performance and efficiency for such demanding recommendation systems.