An Improved Hybrid and Knowledge Based Recommender System for Accurate Prediction of Movies

Recommender system is an adaptive technology and tool that is used in business organizations for offering the products and services by observing their interest and popularity of products. In this paper, an improvement over the existing hybrid and knowledge based recommender system is proposed by integrating the clustering method within content based filter and classification method within collaborative filter. The proposed method handled the scalability problem by using the fuzzy clustering method. This reduced dimension based dataset is processed by the probabilistic Bayesian network classifier for predicting the recommendations. The sparsity problem is handled in both stage of this model. The proposed recommender system model is applied on MovieLens dataset. The comparative analysis was done against content-based recommender system (CBRS), Pearson correlation based collaborative recommender system (PCRS), Frequency-weighted Pearson Correlation (FPC), Weighted Pearson Correlation (WPC) and hybrid recommender systems (HRS). The average RMSE rate achieved by CBRS, PCRS, FPC, WPC, HRS and the proposed hybrid recommender system are 0.3851, 0.3515, 0.3527, 0.3539, 0.3340 and 0.1987 respectively. The significant reduction in MAE rate is also identified in this work. The experimentation results identified that the proposed model reduced the error rate and improved the accuracy rate over existing systems.