Application of SVM model based on collaborative filtering hybrid algorithm in e-commerce recommendation

Abstract

ABSTRACT The current e-commerce suggestion suffers from information overload, which makes it harder for customers to choose products. Due to this, researchers created a hybrid collaborative filtering algorithm and support vector machine e-commerce recommendation model to improve user recommendations. The support vector machine for effective offline classification also explains how to pick the right label threshold, the sparsity of the dataset, and other challenges from both the user- and item-oriented perspectives. According to the experimental results, the support vector machine classification methods based on user perspective and items from the Movielens dataset had average accuracies of 0.91 and 0.88, respectively, while the corresponding average accuracies on the Jester-joke dataset were 0.74 and 0.75. Both the user- and item-based collaborative filtering algorithms on the popular goods dataset generated noticeably fewer mean absolute errors than the cold goods dataset. The average coverage and accuracy of recommendations made using an e-commerce recommendation model that incorporates support vector machines and collaborative filtering algorithms are 90.2%, 92.8%, 88.4%, 83.7%, and 80.2%, respectively. This showed that the research methodology produces positive results in e-commerce recommendation.