Research Article

An Empirical Study of Software Metrics Diversity for Cross-Project Defect Prediction

Table 5

Comparison of the CPDP performance in terms of F-measure.

MetricsTarget
EclipseEquinoxLuceneMylynPde

Semantic0.330 (+62.62%)0.385 (+55.93%)0.450 (+0.22%)0.442 (+12.06%)0.454 (+4.01%)
PS0.415 (+29.22%)0.528 (+13.63%)0.382 (+18.06%)0.379 (+30.61%)0.380 (+24.21%)
OS0.448 (+19.73%)0.566 (+6.17%)0.408 (+10.54%)0.370 (+33.78%)0.413 (+14.29%)
SS0.511 (+4.93%)0.584 (+2.83%)0.451 (—)0.479 (+3.34%)0.453 (+4.19%)
OSS0.5360.6010.4510.4950.472
POSS0.489 (+9.58%)0.583 (+3.07%)0.424 (+6.37%)0.472 (+4.87%)0.436 (+8.26%)

Note: The percentages within the parentheses are the improvement rates of CPDP-OSS, and the boldfaced numbers represent the best results in the same context.