Hybrid Novel Bat Twin Bounded Support Vector Machine Based Personalized Movie Recommendation System

Abstract

The growing demand for personalized content delivery has made recommendation systems a fundamental component in online streaming platforms. The user experience has been significantly enhanced by personalized movie recommendation systems that suggest relevant movies based on individual preferences. In this paper a collaborative filtering based personalized movie recommendation system is proposed that incorporates the novel bat algorithm (NBA) and twin bounded support vector machine (TBSVM), termed hybrid novel bat twin bounded support vector machine (HNBTBSVM). Traditional recommendation approaches are collaborative filtering (CF) and content-based filtering (CB), which often face limitations such as scalability, sparsity, and cold start problems, which limit the system's ability to provide recommendations for new users or items due to insufficient data. A support vector machine (SVM) is a classifier that discriminates between user preferences. The proposed HNBTBSVM model addresses the cold start issue by combining NBA optimization with TBSVM classification. The hybrid model improves classification accuracy and scalability in sparse data environments, a characteristic of cold-start situations. The TBSVM dual-boundary approach effectively handles data sparsity, enhancing boundary flexibility and classification accuracy with limited data. Additionally, similarity measures align new users or items with existing profiles, while augmented data enriches user interactions. Experimental results of the Movielens dataset achieving the accuracy of TBSVM is 85.3%,standard bat algorithm with twin bounded support vector machine(BATBSVM) is 88.3%,contraction factor and dynamic adaptive inertia weight particle swarm optimization support vector machine(CF-IWA PSO-SVM)model is 79.6%,linear kernel and Gaussian kernel HNBTBSVM models are 90.70% and 93.45%,with marked improvements in accuracy, personalization, and cold start mitigation, and reduced the root mean square error(RMSE) and mean absolute error(MAE).