TopK Movie Recommendation Using Matrix Factorization Methods

Recommender systems play a main function in state-of-the-art E-commerce industry. Recommender structures endorse objects to customers which includes books, movies, videos, digital merchandise and lots of different merchandise in general. Recommender systems assist the customers to get customized recommendations. Neighborhood-based Collaborative Filtering (CF) techniques has limitation of scalability and sparseness. To overcome these limitations of CF, model-based matrix factorization is used. Singular Value Decomposition (SVD), Singular Value Decomposition++ (SVD++) and Non Negative Matrix Factorization (NMF) are representative methods of matrix factorization. These methods find the latent factors in the data to find true relationships among different data. Latent factors are Hidden Factors unseen in the data set. This paper proposes the effect of latent factors on SVD, SVD++ and NMF methods for Movie Recommender System. Famous MovieLens dataset from grouplense is used for the implementation of these matrix factorization algorithms and various accuracy metrics like mean absolute error, root mean squared error, precision@k, recall@k and F1 score are calculated for Movie Recommender System.