Research Article
AndroClass: An Effective Method to Classify Android Applications by Applying Deep Neural Networks to Comprehensive Features
Table 9
Classification accuracy (%) of AndroClass for all categories with manual dataset.
| | | | accuracy | | category | #APP | KNN | NB | SVM | DNN |
| 1 | Weather | 15 | 53.3 | 60.0 | 53.3 | 66.6 | 2 | Food & Drink | 21 | 66.7 | 71.4 | 76.1 | 85.7 | 3 | Web Browser | 7 | 42.8 | 57.1 | 71.4 | 71.4 | 4 | Office & Business | 31 | 48.3 | 67.7 | 70.9 | 67.7 | 5 | Keyboard | 10 | 40.0 | 60.0 | 60.0 | 60.0 | 6 | Music & Video | 51 | 49.0 | 80.3 | 88.2 | 90.1 | 7 | Tools | 83 | 61.4 | 71.0 | 81.9 | 91.5 | 8 | Photo & Art | 42 | 57.1 | 64.2 | 76.1 | 78.5 | 9 | Theme & Wallpaper | 37 | 59.4 | 67.5 | 75.6 | 89.1 | 10 | Finance | 15 | 46.6 | 73.3 | 66.7 | 80.0 | 11 | Virtual Reality | 8 | 37.5 | 62.5 | 62.5 | 62.5 | 12 | SNS & Communication | 44 | 54.5 | 70.4 | 77.2 | 88.6 |
| | total accuracy | 55.5 | 70.6 | 76.8 | 83.5 |
|
|