Discrimination Analysis for Predicting Defect-Prone Software Modules

Table 4

Statistical 𝐹-measure values of six classifiers on all data sets. The line 𝑤/𝑡/𝑙 means that the algorithm at the corresponding KDC wins in 𝑤 data sets, ties in 𝑡 data sets, and loses in 𝑙 data sets, compared with the algorithm at the corresponding column.

Project	J4.8	Naive Bayes	RF	AdaBoost	Smote	KDA	LDC
ar3	0.500	0.667	0.633	0.377	0.455	0.702	0.500
ar4	0.474	0.514	0.487	0.443	0.457	0.560	0.485
ar5	0.625	0.667	0.490	0.464	0.575	0.650	0.533
ar6	0.105	0.357	0.254	0.237	0.203	0.201	0.087
pc1	0.271	0.344	0.302	0.347	0.393	0.470	0.031
pc2	0.110	0.055	0.140	0.101	0.100	0.150	0.105
pc3	0.302	0.262	0.357	0.340	0.401	0.436	0.121
pc4	0.503	0.418	0.401	0.437	0.423	0.450	0.342
mw1	0.195	0.390	0.367	0.220	0.283	0.357	0.340
kc2	0.522	0.511	0.474	0.530	0.517	0.650	0.492
kc3	0.357	0.406	0.306	0.328	0.336	0.433	0.433
cm1	0.293	0.256	0.277	0.244	0.217	0.397	0.05
𝑤/𝑡/𝑙	10/0/2	8/1/3	10/1/1	6/4/2	7/5/0	—	10/0/2