In the rapidly evolving domain of recommender systems, a variety of algorithms have been proposed and tested for efficiency, accuracy, and scalability. In the field of recommendation systems, the three main types are content-based filtering, collaborative filtering and the hybrid approach. This technical review delves deep into the implementation and comparative analysis of three prominent algorithms, which correspond to the three types of recommendation systems: K-Nearest Neighbors (KNN), Matrix Factorization, and Deep Neural Networks (DNN). Using two distinct datasets—MovieLens, centered around movie recommendations, and KKBox Music, focusing on musical content—this study aims to furnish insights into the strengths, weaknesses, and nuances of each algorithm's performance. This review not only offers a systematic implementation of the algorithms but also provides insights and to guide the optimal selection and fine-tuning of algorithms in real-world recommender system applications.
- Content-based: K-Nearest Neighbors (KNN)
- Collaborative Filtering: Matrix Factorization (MF)
- Hybrid: Deep Neural Networks (DNN)