<p>Result comparison between Muzaffar et al. [26] and our reimplementation, showing the relative performance of random forest models trained using API usage feature sets reduced using different feature selection algorithms.</p>

For comparison, the corresponding results without feature selection (from Table 11) are also shown.