Evaluation of Classifiers to Enhance Model Selection

Abstract— The various tasks like classification, clustering and association rule deriving are performed in the data-mining for the pattern extraction. The performance evaluation measures make each task distinct and meaningful. The plenty of machine learning algorithms helps in the different ways. The classification helps to predict about the future well in advance and make necessary actions thus it otherwise called as actionable data mining. In this paper we plan to give the overview about various classification algorithms by Waikato Environment for Knowledge Analysis otherwise shortly called as WEKA. The measures found in this helps to determine the best model and proposed statistical analysis namely the paired t-test to enhance the model selection. The evaluations make the promising environment for the model selection. KeywordsEvaluation; Accuracy; T-Test; Data Mining; Classification; WEKA; Stratified Cross Validation; ROC