Skip to main content

Table 15 Performance evaluation of the machine learning classifiers trained on Dataset1 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 0.926 0.919 0.927 0.933 0.9267
MLP 0.902 0.902 0.911 0.911 0.9067
SL 0.897 0.899 0.908 0.906 0.9027
SMO 0.88 0.912 0.917 0.887 0.8991
J48 0.89 0.894 0.903 0.9 0.8972
PART 0.884 0.903 0.91 0.893 0.8834
NB 0.81 0.63 0.72 0.866 0.7493
State-based approach
RF 0.953 0.925 0.934 0.959 0.9427
MLP 0.93 0.921 0.929 0.937 0.9292
SMO 0.918 0.923 0.93 0.925 0.9241
J48 0.918 0.905 0.915 0.927 0.9167
PART 0.91 0.916 0.923 0.917 0.9167
SL 0.905 0.918 0.925 0.913 0.9157
NB 0.741 0.843 0.836 0.732 0.7843
Hybrid approach
RF 0.948 0.911 0.922 0.954 0.9337
MLP 0.937 0.911 0.921 0.944 0.9282
SMO 0.921 0.92 0.928 0.928 0.9247
J48 0.92 0.912 0.921 0.928 0.9202
PART 0.902 0.92 0.926 0.909 0.9147
SL 0.912 0.906 0.915 0.92 0.9137
NB 0.746 0.814 0.816 0.748 0.7795