An Ensemble Classifier Adopting Random Subspace Method based on Fuzzy Partial Mining

Objectives: Ensemble classification with fuzzy partial mining is a novel approach. The random subspace ensemble classifier contains several classifiers working on original attribute space. The aim of this paper is to examine the appropriateness of the random subspace ensemble method for fuzzy partial mining classification and thereby develop an algorithm Ensemble Classification on Fuzzy Utility Mining (ECFUM) by using a skill utility measure in addition to Support and Confidence. Methods/Statistical Analysis: The algorithm show high accuracy with ensemble classifier than solitary classifiers. The classifier is trained on random subspace method which is suitable when there is more number of attributes for the classification, where in many of the fuzzy rule based classification systems suffer increase in dimensionality. Findings: The unique integration of ensemble classification with fuzzy partial weighted mining generates Fuzzy Association Rules and Class Association Rules. Fuzzy association rules have been generated which holds the attributes association. Class association rules have been generated which holds the target class for the attribute association. The resultant classifier produced, shows credible results with better accuracy. ECFUM generates more number of hidden interesting rules compared to traditional associative classifiers. These hidden rules play a major role in later prediction of the algorithm. Improvements: Future work concentrates on the role of infinite sampling on class association rule with higher order confidence precedence to standardize the predictive power of the algorithm ECFUM.