A Hybrid Recommendation System using Collaborative Filtering with Apache Spark in E-Commerce

In the area of product recommender systems, a collaborative recommender system built on a latent factor model has shown notable success. However, there are issues associated with sparsity and its capacity to extract non-linear data features is restricted, which leads to inadequate performance. To overcome the aforementioned challenges, the proposed user-based collaborative filtering creates a social user model for sharing rating patterns to recommend products. The proposed model utilizing Spark architecture and winged hawk optimization (WHO) optimized attention-based CNN is employed for product recommendation. To understand user interests and internet activity, correlation relationships are established using a graph embedding approach, which supports appropriate recommendations for users on e-commerce platforms. The proposed approach on evaluation using the music dataset attains lowest MAE, MSE, and RMSE values as 0.51, 0.55, and 0.74 respectively. The suggested model attained high robustness minimized the computational burden and outperformed other existing models.