Table 3 Data imputation results of detector A in RM missing mode (mean ± std).

From: A hybrid model for missing traffic flow data imputation based on clustering and attention mechanism optimizing LSTM and AdaBoost

Missing rate

10%

20%

30%

40%

50%

60%

HA

45.34

50.19

44.25

47.81

46.38

46.68

33.56

37.32

33.91

35.44

34.37

34.84

18.15

15.71

13.41

14.54

13.60

14.39

KNN

16.50

28.19

26.35

27.50

22.26

24.26

9.79

20.17

18.35

20.60

16.21

18.11

22.27

22.76

16.76

21.13

15.84

16.02

SVM

17.32

21.54

23.64

25.34

26.16

26.64

12.53

15.02

17.46

19.91

20.80

20.38

11.26

13.67

15.51

16.65

17.77

17.39

LRTC-TNN

14.41 ± 2.94

17.81 ± 3.24

19.01 ± 3.15

22.05 ± 3.48

22.47 ± 1.43

25.13 ± 3.53

11.41 ± 2.38

14.06 ± 2.74

15.18 ± 2.62

16.21 ± 2.94

17.31 ± 1.05

20.17 ± 3.05

10.74 ± 1.64

12.66 ± 1.93

13.66 ± 1.75

15.14 ± 2.01

14.57 ± 0.97

17.11 ± 2.34

SDAE

13.21 ± 2.52

16.23 ± 3.30

18.39 ± 3.28

20.53 ± 3.35

24.01 ± 3.94

22.78 ± 3.42

11.14 ± 1.52

12.34 ± 1.63

13.74 ± 1.59

16.01 ± 1.77

18.07 ± 2.01

17.04 ± 1.82

10.51 ± 1.02

12.20 ± 1.21

12.88 ± 1.19

14.25 ± 1.36

15.99 ± 1.67

15.10 ± 1.59

BLSTM-I

11.52 ± 1.91

16.12 ± 2.19

18.11 ± 2.33

19.82 ± 2.78

21.53 ± 2.92

22.24 ± 3.13

9.32 ± 1.02

12.03 ± 1.32

13.11 ± 1.52

14.77 ± 1.67

16.05 ± 1.82

16.86 ± 2.01

9.59 ± 1.09

11.06 ± 1.21

12.55 ± 1.41

13.54 ± 1.49

14.09 ± 1.52

14.51 ± 1.72

MDGCN

10.99 ± 1.78

13.02 ± 2.19

14.31 ± 2.41

15.18 ± 2.51

15.47 ± 2.41

16.42 ± 2.51

8.74 ± 1.28

9.91 ± 1.34

10.92 ± 1.52

11.32 ± 1.62

11.53 ± 1.57

12.53 ± 1.84

7.81 ± 1.03

9.53 ± 1.25

9.72 ± 1.33

10.17 ± 1.44

10.27 ± 1.39

10.72 ± 1.74

Proposed model

8.09 ± 1.12

9.31 ± 1.23

10.83 ± 1.43

11.23 ± 1.55

11.57 ± 1.59

13.07 ± 1.71

6.06 ± 0.63

6.99 ± 0.73

7.21 ± 0.82

7.28 ± 0.99

8.27 ± 1.21

8.59 ± 1.33

4.91 ± 0.52

4.83 ± 0.55

5.32 ± 0.62

5.56 ± 0.67

5.60 ± 0.69

5.91 ± 1.01