An Optimized Approach for Feature Extraction in Multi-Relational Statistical Learning

Abstract

Various features come from relational data often used to enhance the prediction of statistical models. The features increases as the feature space increases. We proposed a framework, which generates the features for feature selection using support vector machine with (1) augmentation of relational concepts using classification-type approach (2) various strategy to generate features. Classification are used to increase the productivity of feature space by adding new techniques used to create new features and lead to enhance the accuracy of the model. The feature generation in run-time lead to the building of models with higher accuracy despite generating features in advance. Our results in different applications of data mining in different relations are far better from existing results.