Sparse Matrix Completion for Effective Recommendation System

In this era of information retrieval, the revenue of the e-commerce system mainly rely upon how intelligently information is being processed and decision are being made. In this paper, the problem of unavailability of complete information is targeted for information processing and decision-making in e-commerce domain. The existing system which works on a similar phenomenon is known as recommendation system which becomes the most evolving subject in the area of electronically operated markets. However, the decision-making capability and suggestive nature of these techniques have improved the overall output of the electronic market, but lacking behind where sufficient information is not available. In literature, most of the existing schemes are designed using collaborative filtering and content-based recommendations with KNN and K-means. But, huge increment in the number of online users and their varied pattern of purchasing goods increase the sparsity in information matrix due to which neighbor selection is getting more problematic. The proposed recommended system embeds the collaborative filtering method to complete the obtained incomplete matrix which lifts the aforementioned problem of high sparsity in data. The proposed scheme is verified and validated over different datasets and achieve significant results over existing schemes.