Skip to main content

An Open Access Journal

Table 3 F-Score based performance evaluation for proposed scenarios

From: Detection of anomalies in cycling behavior with convolutional neural network and deep learning

Scenario A1: Training = 80% of dataset, Testing = 20% of dataset, Preliminary Selection: Filtered data with TS_window size 40

Scenario A2: Training = 100% dataset of any one user, Testing = 100% dataset of all other users, Preliminary Selection: Filtered data with TS_window size 40

User

T

RP

CNN

Performance metrics

User’s for training

T

RP

CNN

Performance metrics

Recall

Precision

F2

Recall

Precision

F2

ID-1

0.21

3

4

1

0.75

0.94

ID-1

0.18

33

47

0.76

0.53

0.70

ID-2

0.21

4

3

0.75

1

0.79

ID-2

0.18

37

39

0.70

0.67

0.69

ID-3

0.21

1

2

1

0.5

0.83

ID-3

0.18

38

47

0.76

0.62

0.73

ID-4

0.21

0

3

-

-

 

ID-4

0.18

41

43

0.78

0.74

0.77

ID-5

0.21

1

1

1

1

1.00

ID-5

0.18

38

46

0.74

0.61

0.71

ID-6

0.21

3

3

1

1

1.00

ID-6

0.18

35

47

0.77

0.57

0.72

ID-7

0.21

1

2

1

0.5

0.83

ID-7

0.18

39

49

0.80

0.65

0.76

ID-8

0.21

2

3

0.5

0.33

0.45

ID-8

0.18

34

41

0.76

0.63

0.73

ID-9

0.21

1

2

0

0

0.00

ID-9

0.18

37

45

0.73

0.60

0.70

ID-10

0.21

2

2

1

1

1.00

ID-10

0.18

37

41

0.65

0.59

0.64

Weighted Average

0.21

18

26

0.83

0.60

0.77

Weighted Average

   

0.75

0.62

0.72

  1. The significance of bold is linked to the best result of the model