Skip to main content

Table 3 Comparison of performance of M5Rules and RNN models on different 2 × 1 dataset combinations

From: Predicting rice blast disease: machine learning versus process-based models

Training data*

Test data

Train

Test

r (a)

r (b)

r (a)

r (b)

MAE** (a)

MAE**

(b)

r2 (a)

r2 (b)

k2015 + k2016

s2016

0.94

0.94

0.56

0.72

0.78

0.63

0.31

0.52

k2015 + k2016

p2015

  

0.21

0.31

0.60

0.88

0.04

0.10

k2015 + p2015

k2016

0.87

0.85

0.69

0.68

0.78

0.53

0.48

0.46

k2015 + p2015

s2016

  

0.67

0.57

0.91

0.71

0.45

0.32

k2015 + s2016

k2016

0.92

0.95

0.46

0.71

0.39

0.51

0.21

0.66

k2015 + s2016

p2015

  

0.22

0.25

1.04

1.10

0.05

0.06

k2016 + p2015

k2015

0.92

0.94

0.61

0.69

0.16

0.48

0.37

0.47

k2016 + p2015

s2016

  

0.52

0.58

0.98

0.69

0.27

0.33

k2016 + s2016

k2015

0.89

0.92

0.36

0.73

0.40

0.52

0.13

0.62

k2016 + s2016

p2015

  

0.17

0.35

1.20

0.78

0.03

0.12

s2016 + p2015

k2015

0.91

0.94

0.33

0.63

0.80

0.63

0.11

0.40

s2016 + p2015

k2016

 

0.19

0.57

0.72

0.62

0.04

0.32

Average

0.42

0.57

0.73

0.67

0.21

0.37

  1. *k = Kalochori, s = Seville, p = Portugal; **Mean absolute error, (a) = M5Rules, (b) = RNN