Hybrid Recommendation Systems using Adaptive Clustering to address Cold start problems
Md. Mijanur Rahman, Fahim Faisal Sifat, Rakibul M. Islam, Safaeat Molla, Md Rezaul Karim Khan
Abstract
This study aims to address the “cold start” problem and solve recommendation systems” current challenges., particularly for new users. The primary goal is to enhance the accuracy and customization of book suggestions in digital libraries and e-commerce sites. Our study uses a hybrid recommendation system., combining content-based and collaborative filtering techniques. Fuzzy user-based clustering enhances collaborative filtering., creating multidimensional user profiles based on user ratings., age., and geographical data. The system architecture illustrated in Figure 1 includes data collection., cleaning., and the creation of a feature-rich dataset. This proposed model handles the complexity of the recommendation system and ensures a systematic approach. By combining collaborative and content-based filtering., this system can solve the cold start problem. The fuzzy c-means clustering plays an essential role in grouping similar users., thus improving new users” accuracy. This system uses a re-recommendation process after user data removal that marks the adaptability of the system. The similarity score uses cosine similarity and evaluates and demonstrates the system's potential to offer customized and accurate book recommendations. This proposed system establishes a new benchmark in the field of recommendation systems. This approach contributes to a more user-centric., intuitive., and responsive approach to book recommendations. By solving cold start problems., our model enhances the user experience in digital reading choice. This study may explore the diverse recommendation scenarios beyond books in future research.