Skip to main content

Table 2 Comparison of Out-of-bag prediction rates (%) from Training sets and Prediction rates from Test sets

From: Application of machine learning to predict transport modes from GPS, accelerometer, and heart rate data

 

Naïve out-of-bag prediction rates in the training sets

Prediction ratesa in the test sets

Before correction

After correction

Before correction

After correction

Overall

94 (94–94)

78 (78–78)

94 (81–98)

79 (62–88)

Overall transport

70 (70–70)

77 (76–77)

64 (33–88)

74 (41–89)

Activity place

98 (98–98)

78 (78–79)

98 (95–100)

79 (61–89)

Bike

67 (66–68)

88 (88–89)

65 (40–82)

90 (77–100)

Private motorized

84 (84–85)

72 (72–73)

85 (16–96)

69 (1–89)

Public transport

46 (44–47)

75 (74–75)

22 (0–68)

62 (0–91)

Walking

63 (62–63)

81 (81–82)

61 (5–88)

80 (23–95)

  1. a Prediction rates presented as median prediction rates from 126 RF models with 2.5th and 97.5th percentiles in the parentheses. Prediction rates by transport modes are shown before and correction for category size; model without heart rate data
Back to article page