Improving the QoS of Recommender Systems using Adaptive Machine Learning

Abstract

Recommendation based systems have gained a lot of popularity due to their wide range of applicability. From e-commerce-based product recommendation, to social media-based friend recommendation, these systems can be used for any kind of pattern analysis targeted to recommending data based on interlinked usage statistics. In order to improve the quality of such systems, they must have a strong pattern recognition engine, combined with a strong prediction engine. Because, a strong pattern recognition engine will be able to analyze and distinguish different patterns effectively, and the prediction engine will be able to merge these patterns together in order to predict the recommendation for the system. Generally, algorithms like neural network, k-means, kNN and SVM based pattern analyzers are combined with neighborhood-based, context-aware pattern analysis-based and collaborative filtering-based predictors in order to develop a complete recommendation system. Many authors have also combined recommender systems in order to generate a high-performance hybrid recommender. In this paper we have proposed a machine learning based recommender system which uses strengths of different recommender systems in order to improve the overall recommendation accuracy. Furthermore, the performance is compared with some state-of-the art systems in order to evaluate the performance of the proposed system