Skip to main content

Table 7 ILI prediction results (current week estimates for %ILI) for 2014 for 31 locations from affect features using Linear, AdaBoost and Random Forest regressors

From: Uncovering the relationships between military community health and affects expressed in social media

 

Linear

AdaBoost

RandomForest

 

CORR

MAPE

RMSE

CORR

MAPE

RMSE

CORR

MAPE

RMSE

L0

0.10

14.5

0.69

0.15

18.2

0.80

0.24

15.40

0.65

L3

0.21

32.0

1.32

0.25

18.4

0.96

0.39

17.90

0.87

L4

0.07

23.4

1.04

0.56

18.9

0.82

0.28

23.60

0.87

L10

0.19

26.9

0.84

−0.60

60.6

1.44

0.09

30.40

0.87

L11

0.22

24.6

1.08

0.13

25.8

1.02

0.18

27.70

0.92

L12

0.10

43.6

1.70

0.08

34.8

1.48

0.12

67.10

2.09

L13

0.28

23.0

1.11

0.28

22.7

1.20

0.21

22.70

1.13

L14

0.19

34.1

1.49

−0.15

37.8

1.20

0.41

28.90

0.92

L15

−0.01

37.7

1.26

−0.01

50.6

1.63

0.26

40.10

1.17

L16

−0.07

26.6

1.30

0.06

24.8

1.29

−0.19

30.50

1.37

L19

0.09

32.1

1.61

0.15

28.1

1.61

−0.28

35.00

1.78

L20

0.16

16.9

1.23

0.28

18.2

1.18

0.21

22.90

1.27

L21

0.06

29.1

1.55

0.25

32.7

1.41

0.03

50.30

1.80

L22

−0.16

29.8

1.62

−0.44

41.1

1.68

0.08

42.70

1.56

L23

0.06

28.3

1.14

−0.12

41.8

1.27

0.06

37.80

1.18

L25

0.24

18.9

0.88

−0.06

28.9

1.11

0.03

24.30

0.98

L27

0.23

32.2

1.57

0.20

22.7

1.24

0.21

23.20

1.20

L28

0.10

22.0

1.39

−0.01

24.0

1.31

−0.21

27.00

1.29

L29

−0.11

29.5

1.67

-0.07

37.0

1.48

−0.16

31.80

1.35

L30

0.16

28.7

1.37

0.11

23.0

1.33

0.23

38.00

1.28

L31

0.12

35.4

1.22

0.13

39.9

1.30

0.11

35.40

1.22

L32

−0.29

25.0

1.45

−0.48

33.7

1.51

0.13

26.00

1.22

L33

0.14

19.6

1.16

0.37

17.3

1.04

0.50

18.00

1.00

L34

0.18

19.8

0.99

−0.10

29.4

1.32

0.16

23.90

1.11

L37

0.09

22.7

1.67

0.29

20.2

1.34

0.49

18.70

1.34

i2

0.12

12.1

0.73

0.28

11.7

0.66

0.13

13.70

0.62

i3

0.33

19.2

0.78

−0.01

25.3

0.95

−0.08

40.20

1.42

i17

0.17

71.0

2.42

−0.09

57.8

1.84

0.15

52.70

1.74

i20

0.09

34.6

1.09

−0.14

38.4

1.21

−0.12

39.30

1.17

i25

0.58

35.1

1.54

−0.03

32.9

1.65

−0.08

33.10

1.50

i27

0.03

20.8

1.67

0.29

23.2

1.44

0.17

21.20

1.49

  1. Models are trained exclusively on emotion and sentiment proportions extracted from tweets in 2012-2013. Models are evaluated to predict %ILI in 2014. The highest CORR and the lowest MAPE and RMSE are highlighted in bold.