Development of Efficient Movie Recommender System for a Group of Users based on Their Preferences

Abstract

Recommender Systems have gained popularity recently due to their ability to expedite users’ selection processes. Traditional recommender systems mainly focus on providing recommendations to a user. It is not suitable for recommending an item to groups of users. A group recommendation system (GRS) addresses this issue of recommendation. Group recommender system is popular in a few domains such as parties, tourism, movies, etc. Movie group recommender system is a special kind of GRS that is designed for movie recommendations to the group of users. In contrast to traditional Recommender Systems (RS), movie GRS has gained prominence for its ability to cater to the collective preferences of groups. The task of GRS is divided into two tasks such as group item preference prediction and group item recommendation. Nowadays, group satisfaction has become a major issue in the area of GRS. Group user satisfaction plays a vital role in collective decision-making in a group. Many researchers have developed various algorithms to address this issue. However, prior GRS techniques failed to address the important issue of group satisfaction. The efficacy of GRS relies heavily on understanding group preferences, encompassing factors like trust, influence, and likeness among group members. An ongoing challenge in the movie GRS pertains to member relationships and group satisfaction. However none of the researchers provide a good user satisfaction with less error rate. This thesis tries to address the issue of group user satisfaction of movie GRS in three contributory chapters. The first contributory chapter uses the collaborative filtering approach and tries to enhance group satisfaction. It explores member inclination and item usefulness within a group to address this issue. The proposed methodology employs an aggregate prediction technique to calculate the final group score. It computes user inclination and item usefulness at both the individual and group levels. A novel aggregation strategy, Popularity and Likeness-based Aggregation (PLAS) is developed to aggregate individual predictions into a complete group score. Another collaborative filtering (CF) approach, revised slope one, is proposed to predict group member preferences. Existing average modelling combines the preference to get a final group preference. Experimental outcomes show the superiority of the proposed method. The second contributory chapter uses the content information with the user profile to improve group satisfaction further. It employs cluster validation metrics for the selection of appropriate clusters and cluster sizes in the group formation stage. Later, a novel technique is proposed for predicting individual member ratings considering user characteristics or genre information to optimize group user satisfaction. Based on user and item characteristics, the newly introduced user inclination and item usefulness-based aggregation function aims to enhance the aggregation process. Later, the study further gives another approach to improve group user satisfaction with less error. The study introduces novel similarity methods to predict group user ratings individually. A linear neural network model is introduced to aggregate individual ratings into a group score effectively. Finally, the performance is evaluated on standard datasets. Experimental outcomes show the superiority of the newly introduced approach. Finally, this study contributes to the evolving field of movie GRS by addressing the challenges of group user satisfaction. Also, this chapter explores the fusion of sequential and group recommendations through the ”Dynamic Group Recommendation using Sequential Recommendation” (DGRSR) method, showing improved group predictions and recommendations in a linear approach. The experiments are performed on the standard datasets. Results show the superiority of the state of art techniques.