Greedy Strategy and Co-Evaluation Factor in Collaborative Filtering: Enhancing Recommender Systems with Potential Applications in IoT

Recommender systems (RS) have emerged as a research hotspot for offering information retrieval services in light of the Internet's tremendous growth in data. Because of their effectiveness and simplicity, collaborative filtering (CF) algorithms have attracted a lot of attention. When user data is exceedingly scarce or when there is no past data for users who are new to the system, CF algorithms' prediction accuracy decreases. A few deep learning-based advancements in algorithms enhance computational complexity while simultaneously improving accuracy. In order to simplify the computing process and address the issue of data sparsity, we combine the greedy technique used in the approximation algorithm with the CF algorithm and create a “co-evaluation” factor in this study. To further ease the cold start issue, we expand the technique by adding user attribute data. Lastly, we run trials on the MovieLens dataset and find that the prediction performance is significantly better than with the conventional CF algorithm. While the study focuses on movie recommendations, the techniques developed hold significant promise for application in the Internet of Things (IoT). In IoT environments, where data from multiple heterogeneous sources must be effectively fused and processed, our approach offers a potential framework for enhancing decision-making processes by efficiently integrating diverse data streams and addressing sparsity challenges.