Pragmatic Method For Improving Accuracy In Recommender Systems

The purpose of recommender systems is to provide suggestions for items to be of use to a user. Recommender systems tools and techniques are now popular both commercially and at the research community, where many ideas have been suggested for providing recommendation by filtering the information backstage to predict whether a user would like a given item. The suggestions provided by Recommender systems are attempt to help the users in their decision-making processes, such as what items to buy, what music to listen, or what news to read. Many techniques to generate recommendations have been invented and during the last decade, many of them have also been employed successfully in commercial environments.A system designer who wishes to exploit recommender systems has to choose between a set of candidate approaches. A first step towards choosing an appropriate algorithm is to decide which properties of the application to focus upon when making this choice. Indeed, recommendation systems have a set of properties that may affect a user experience, such as accuracy, robustness, scalability, and so forth.Collaborative Filtering (CF) and Content-Based Filtering (CBF) recommender systems suffer from potential problems, such as sparsity, reduced coverage, cold-start, and overspecialization, which reduce the effectiveness of these systems. Hybrid recommender systems combine individual recommendation approaches to overcome some of the aforementioned problems. We propose novel switching hybrid recommendation algorithms using classification approaches trained on the content profiles of users and item-based CF. A switching hybrid recommender system is intelligent in the sense that it can switch between recommendation approaches using some criteria. The benefit of a switching hybrid hybrid recommender is that it can make efficient use of strengths and weaknesses of its constitutional recommender systems, our recommendation algorithms give more accurate results than the conventional hybrid approaches. We show empirically that the proposed algorithms outperform (or give comparable results to) other recommender system algorithms in terms of the MAE, ROC-Sensitivity, and coverage; while at the same time eliminate some of the recorded problems with recommender systems. Furthermore, they maintain robust performance under the cold-start scenarios. We evaluate our algorithm over the MovieLens (SML) and FilmTrust (FT1) datasets.By switching between the machine learning classifiers and the CF approach, our algorithms can balance the accuracy and diversity of recommendations. If we construct a list of top-N recommendations for an active user, then our algorithms would introduce some sort of randomness in the recommendation list, resulting in a range of alternatives to be recommended rather than a homogeneous set of items.The evaluation we have explain in this work using MAE, ROC-sensitivity, and coverage of different algorithms shows that the proposed algorithms outperform others significantly in terms of the MAE and ROC-sensitivity; whereas they give comparable results to others in terms of coverage metric. It’s also shows that the proposed algorithms are scalable and practical as their on-line cost is less than or equal to the cost of other algorithms. We are using the item-based CF, whose on-line cost is less than that of the user-based CF.