Comparing the Performance of Data Mining Tools: WEKA and DTREG

The objective of the paper is to compare two data mining tools on the basis of various estimation criteria. The data mining tools which are evaluated are WEKA and DTREG. These tools are used to build multilayer perceptron which is a data mining model to predict the survivability of the oral cancer patients. Oral cancer database is considered as it is estimated to be 8th most common cancer worldwide and extremely grave problem in India as well. Early detection is the only way to prevent the disease and reduce this burden. Dtreg is a proprietory data mining tool whereas weka is an open source. Classification accuracy of multilayer perceptron model developed using dtreg is 70.05% and using weka is 59.70%. 10-fold cross-validation method is used for validation by dtreg and stratified cross validation is used by weka. The data mining tool dtreg has demonstrated better results in terms of true negative, false negative, specificity, recall and area under ROC curve. However, weka displays better results in terms of true positive, false positive, precision and f-measure. Analysis run time of dtreg is less than weka and the report generated by dtreg is also more expressive and descriptive in comparison to weka, which makes dtreg a better data mining tool for multilayer perceptron models.