An Efficient Algorithm for Recommender System Using Kernel Mapping Techniques

Abstract

Recommender Systems is a system that helps users to find good stuff like movies, books etc. The user gives ratings to items and it is a system that predicts these ratings. User rates those items that are of their interest. Two types of techniques are used by the recommender systems for recommendations-content based filtering (CBF) and collaborative filtering (CF). Both techniques have their own pros and cons. The most common problems with CF are a cold star, scalability, and sparsity. Also, collaborative filtering needs a large amount of data. We will propose a solution using Kernel Mapping Recommender (KMR) to resolve the recommendation issues like a cold star, scalability, and sparsity the main goal of proposed work is to find dynamic, efficient and effective recommendation algorithm that can be successfully used to make a recommendation to users'. Several heuristic algorithms have been anticipated that merge different categories of KMR for improving correctness and removal of problems linked with a predictable recommender system. The proposed system is checked on movie datasets that are available online and then standardized with KMR. In conditions of accuracy, precision, recall, F1 measure, and ROC metrics; the results expose that the proposed algorithm is quite precise mainly under cold-start and sparse situations.