A Comparative Study in Recommendation Systems
Click the Poster to View Full Screen, Right click to save image
Grant: CURF
Noorul Sama Sahel
College:
The Dorothy and George Hennings College of Science, Mathematics, and Technology
Major:
Computer Science
Faculty Research Advisor(s):
Daehan Kwak
Abstract:
Recommender systems play a crucial role in various fields such as e-commerce and entertainment platforms. They improve user experience and engagement by offering users personalized suggestions. A comparison of two well-known recommendation techniques is presented in this abstract: Collaborative Filtering (CF) and Content-Based Filtering (CBF). Collaborative filtering, a widely used technique, leverages user-item interactions to generate personalized recommendations. It analyzes user interactions with items, such as ratings or purchase history, to identify users with similar preferences. The system then recommends items that those users have enjoyed but the target user has not yet explored. However, CF may face challenges such as the cold-start problem for new users or items with limited ratings, and scalability issues with large datasets. On the other hand, Content-Based Filtering (CBF) provides individualized recommendations that are in line with user preferences by suggesting items based on their characteristics and attributes. In the context of book recommendation, CBF analyzes the content or attributes of books themselves, such as genre, author, writing style, themes, and plot elements. This comparison study examines several aspects of CF and CBF in book recommendations, such as scalability, handling of the cold start problem, recommendation accuracy, and data needs. Although CF is great at gathering user preferences and encouraging unexpected discovery, CBF addresses the cold start problem and guarantees relevance by emphasizing book qualities. Moreover, the study looks into the possibilities of hybrid recommendation strategies that combine CF and CBF approaches to capitalize on the advantages of each strategy. Hybrid systems strive to provide more varied and reliable book suggestions by utilizing content-based and collaborative insights, taking into account the wide range of reader interests. In conclusion, a hybrid strategy that combines the benefits of exploratory data analysis and collaborative filtering shows potential for improving book recommendation systems. Such hybrid systems can provide users with recommendations that are more accurate, relevant, and appealing by combining individualized recommendations with extensive dataset insights. Further research and experimentation are needed to examine how these approaches can work together and create reliable recommendation systems for a range of datasets and user preferences.