DNN based movie recommendation system with explainable reasoning

Abstract

A recommender system gives confidential private preferences to its users by evaluating users’ historical data and behaviors. Since the early 1980s, the recommender system has been used as a popular symbolic AI technology. It comes under the broad umbrella of the Expert system. However, with the latest developments, The use of machine learning techniques and algorithms in recommender system development has been a natural way to increase prediction accuracy and address issues with data sparsity and cold start. Users and service providers can both benefit from recommender systems. Frequently, visitors are directed to recommended articles that they were not expecting, which can lead to confusion. As a result, explainability becomes to be evaluated in direct relation to how well the recommendation system operates. Enhancing efficiency, transparency, and customer happiness leads to increased user loyalty. The recommendations are supported by justifications for the approaches. Hence, with the rapid growth of the demand for recommender systems, its landscape is evolving into explainable recommender systems for better quality. Since then explainable recommender systems become a research challenge. Some techniques for generating explanations of recommendations are provided by the academia. The Recommender system is depending more and more on these machine learning algorithms due to the recent success of machine learning models and algorithms. Explainable Artificial Intelligence is a new field that aims to produce high-quality suggestions and clear explanations of the outcomes while offering transparent learning techniques. This has opened a new era of recommender systems, revealing more complex relationships between users and products, offering highly sophisticated data visualizations, and discovering vast amounts of information in contextual, virtual, demographic, and textural data. Therefore, this thesis reports on the development of a machine learning-based recommender system solution with explainable reasoning.