Skip to main content

Table 16 Performance evaluation of the machine learning classifiers trained on Dataset2 using random-based, state-based and hybrid approaches

From: Machine learning-based dynamic analysis of Android apps with improved code coverage

Classifier PrecM RecM PrecB RecB W-FM
Random-based approach
RF 0.855 0.835 0.804 0.827 0.8319
PART 0.828 0.805 0.77 0.796 0.8011
J48 0.819 0.79 0.754 0.787 0.7887
MLP 0.825 0.741 0.719 0.809 0.7720
SMO 0.733 0.795 0.721 0.647 0.7265
SL 0.731 0.794 0.719 0.643 0.7241
NB 0.741 0.536 0.577 0.772 0.6391
State-based approach
RF 0.904 0.87 0.848 0.887 0.8774
PART 0.872 0.84 0.813 0.849 0.8447
J48 0.862 0.828 0.8 0.838 0.8333
MLP 0.868 0.814 0.789 0.85 0.8301
SMO 0.866 0.759 0.745 0.856 0.8031
SL 0.847 0.772 0.749 0.83 0.7985
NB 0.744 0.665 0.639 0.721 0.6917
Hybrid approach
RF 0.891 0.864 0.84 0.871 0.8671
PART 0.862 0.835 0.806 0.838 0.8358
J48 0.848 0.825 0.793 0.82 0.8229
MLP 0.842 0.831 0.797 0.81 0.8217
SL 0.837 0.773 0.747 0.816 0.7932
SMO 0.846 0.754 0.735 0.833 0.7898
NB 0.732 0.646 0.622 0.711 0.6761