An effective data classification method for medical dataset in terms of accuracy and time

Abstract

Data classification plays a major role in organizing the optimal features for the most effective and efficient use. Feature Selection technique is one of the foremost methods to select the optimal features from the dataset. The classification accuracy and the processing time required to build the model are the two main keys in obtaining the effective data classification by using the feature subset methods and ranking methods. The work was tested on seven real time dataset (Breast Cancer, Breast Tissue, Contact Lenses, Dermatology, Hypothyroid, Iris and Liver Disorders) obtained from UCI Data repository. The results obtained from CFS Subset Attribute Evaluator, Correlation Attribute Evaluator, Gain Ration Attribute Evaluator, Info Gain Attribute Evaluator, OneR Attribute Evaluator, Principal Components Attribute Evaluator, ReliefF Attribute Evaluator, Symmetrical Uncertainty Attribute Evaluator and Wrapper Subset Attribute Evaluator were compared. Classification algorithms like Navis Bayes, Bayes Net, Multilayered Perception, Sequential Minimum Optimization, K Nearest Neighbours, Decision Tree, OneR, J48 and Random Tree are used to analyze the classification accuracy and processing time. Comparison are done with the results obtained by using the ranking methods and the results obtained bynot using the ranking method, to find whether the ranking methods are important in obtaining the classification accuracy and processing time.