A Novel Movie Recommendation System with Iterated Truncated Singular Value Decomposition (ITSVD)

Recommendation systems are one of the most essential tools for e-commerce intelligence. These systems with different types of data filtering methods are able to offer the best recommendations from a multitude of selectable items. Collaborative Filtering is the most widely used method of filtering data to make recommendations. One of the advanced models for predicting ratings in the Collaborative Filtering is the Singular Value Decomposing (SVD). In this paper, an optimized model of the film recommending system based on the SVD method is developed, which while reducing the dimensions of the matrices and the volume of computations and memory, and with iteration replacement method, has appropriate accuracy compared with other methods. For this research, a set of 100k Movie Lens datasets and Python programming have been used. Evaluation of error rate with root mean square error (RMSE) and mean absolute error (MAE) value shows a good improvement over similar methods in other references.vv