Table 4 The overall test accuracy of predicting the cluster indices using CNN4 compared against the total accuracy using regular logistic regression algorithm (Log-Reg).

From: Predicting clustered weather patterns: A test case for applications of convolutional neural networks to spatio-temporal climate data

Lead days

CNN4 Summer

CNN4 Winter

Log-Reg Summer

Log-Reg Winter

Lead day 1

92.1% ± 0.4%

93.3% ± 0.2%

65.3% ± 0.2%

66.7% ± 0.3%

Lead day 2

89.3% ± 0.6%

91.1% ± 0.3%

63.2% ± 0.4%

65.8% ± 0.7%

Lead day 3

83.4% ± 0.2%

87.2% ± 0.2%

59.7% ± 0.2%

63.1% ± 0.3%

Lead day 4

80.1% ± 0.3%

82.4% ± 0.1%

56.8% ± 0.7%

63.2% ± 0.4%

Lead day 5

76.4% ± 0.4%

80.3% ± 0.2%

55.4% ± 0.2%

61.7% ± 0.3%

Back to article page