Adaptive and Nonlinear Strategies based Switching Hybrid Model for Book Recommendation using User Activity and Data Availability

Abstract

The diverse and constantly evolving preferences of the users require the library management system to always meet them. Therefore, in this digital era, Book Recommendation Systems online, in particular, have been the key to making it easy for the users by offering personalized book recommendations, thus raising user engagement as well as overall satisfaction. In this regard, we discuss major developments in recommendation methods, with emphasis on collaborative, content-based, and combined methods. In this paper, we introduce a novel hybrid recommendation model that dynamically switches between two well-established methods: K-Nearest Neighbors (KNN), and Neural CF (NCF). The main purpose of our approach is the hybridization of the strengths from two models. As observed, the KNN-based CF offers good interpretability with the relative ease, while the NCF is capable to retrieve the nonlinear interactions between users and items. Unlike other hybrid approaches which combine the outputs of multiple models, the proposed method follows a dynamic switching approach. This allows the system to select the best model based on specific user conditions to improve the recommendation process. In addition, we propose an enhanced version of this algorithm that makes use of expert users to optimize the precision of the recommendation system. The performance of the proposed models is measured in terms of metrics such as RMSE, MAE, precision, and recall, and it is observed that the proposed models outperform the baseline algorithms.