Efficient Feature Selection Method Using Feature Grading for Context-Aware Recommendation Systems

Abstract

In e-commerce industry, with rapid growth an importance was gained by recommendation system. To suggest products, users feedback was utilized by the recommendation system, that help in accessing the long tail products and also might be useful to the user. To get the correct results from database, it is very necessary to retrieve correct data, it is very big and difficult task to handle these large dataset. In a dataset, feature selection is the automatic attributes selection, which is known as attributes or variable selection, to the predictive modeling problem it is most relevant. Importance of item features makes recommendations more relevant. Therefore, this paper presents Efficient Feature Selection Method Using Feature Grading for Context-Aware Recommendation Systems. In e-commerce business, to make commodities recommendation, a comprehensive algorithm is used which is known as Feature-Grading. Feature-Grading process as follows: Extracting overall feature set commodities of a group category, in the original data, to select and rank the relevant features in binary space, the important features density and weights are considered and finally acquiring feature set based subset of grading set. Root-mean-square-error (RMSE) is the metric for measuring the prediction accuracy. Feature-Grading really works well or not will be revealed by some important results which are discussed by our experimental data.