Table 4.

Comparison of the performance of different models using the combined dataset

Classification Algorithms	Percent Correctly Classified	AUC	Precision	Recall	F-Measure
AdaBoost	84.28%	0.849	0.981	0.484	0.649
Bayes Net	81.55%	0.889	0.674	0.744	0.707
Decision Stump	84.31%	0.730	1	0.476	0.645
Decision Table	84.29%	0.865	0.923	0.519	0.665
J48a	84.64%	0.843	0.97	0.503	0.662
JRip	84.27%	0.737	0.985	0.483	0.648
LMT	85.12%	0.901	0.867	0.594	0.705
Logistic	85.06%	0.899	0.835	0.626	0.715
Native Bayes	82.63%	0.862	0.768	0.603	0.675
OneR	84.88%	0.749	0.993	0.499	0.664
PART	84.58%	0.869	0.973	0.500	0.660
Random Forest	84.60%	0.893	0.949	0.514	0.667
Random Tree	76.77%	0.776	0.628	0.552	0.588
REP Tree	83.61%	0.845	0.884	0.522	0.656
SGD	84.90%	0.765	0.904	0.555	0.688
Simple Logistic	85.12%	0.901	0.867	0.594	0.705
SMOb	84.84%	0.751	0.972	0.509	0.668
Voted Perceptron	70.70%	0.519	0.773	0.031	0.06
Deep FNN	82.02%	0.751	0.831	0.820	0.813

^aThe implementation of C4.5 decision tree in WEKA

^bWEKA’s implementation of John Platt's sequential minimal optimization algorithm for training a support vector classifier.