A comparison of data mining methods for prediction and classification types of quality problems

Thesis Type: Postgraduate

Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Faculty of Engineering, Department of Industrial Engineering, Turkey

Approval Date: 2009




In this study, an Analytic Network Process (ANP) and Preference Ranking Organization MeTHod for Enrichment Evaluations (PROMETHEE) based approach is developed and used to compare overall performance of some commonly used classification and prediction data mining methods on quality improvement data, according to several decision criteria. Classification and prediction data mining (DM) methods are frequently used in many areas including quality improvement. Previous studies on comparison of performance of these methods are not valid for quality improvement data. Furthermore, these studies do not consider all relevant decision criteria in their comparison. All relevant criteria and interdependencies among criteria should be taken into consideration during the performance evaluation. In this study, classification DM methods namely; Decision Trees (DT), Neural Networks (NN), Multivariate Adaptive Regression Splines (MARS), Logistic Regression (LR), Mahalanobis-Taguchi System (MTS), Fuzzy Classifier (FC) and Support Vector Machine (SVM); prediction DM methods DT, NN, MARS, Multiple Linear Regression (MLR), Fuzzy Regression (FR) and Robust Regression (RR) are prioritized according to a comprehensive set of criteria using ANP and PROMETHEE. According to results of this study, MARS is found superior to the other methods for both classification and prediction. Moreover, sensitivity of the results to changes in weights and thresholds of the decision criteria is analyzed. These analyses show that resulting priorities are very insensitive to these parameters.