Skip to main content

Table 2 (a) The RMSE, MAE and ICC statistics of the used methods for prediction of ILI; (b) the performance criteria of the used methods for prediction of ILI outbreaks

From: Comparative evaluation of time series models for predicting influenza outbreaks: application of influenza-like illness data from sentinel sites of healthcare centers in Iran

(a)
Model Kernel   Criterion
RMSE MAE ICC
RFTS Train 25.3 6.43 0.92
Test 22.78 14.99 0.88
SVM RBF Train 58.71 14.3 0.58
Test 28.19 22.36 0.53
Polynomial Train 55.20 15.00 0.53
Test 239.00 91.20 0.09
Linear Train 53.60 13.00 0.53
Test 30.10 18.60 0.47
Sigmoid Train 63.90 17.30 0.43
Test 30.80 20.00 0.24
ANN Train 37.50 11.94 0.84
Test 26.58 13.21 0.82
ARIMA Train 47.01 17.92 0.64
Test 34.90 28.16 0.03
(b)
Model Sensitivity Specificity Criterion
PPVa NPVb Total accuracy
RF
 Train 1.000 1.000 1.000 1.000 1.000
 Test 0.804 0.964 0.974 0.750 0.865
SVM
 Train 1.000 1.000 1.000 1.000 1.000
 Test 0.848 0.964 0.975 0.794 0.892
ANN
 Train 0.962 0.940 0.828 0.977 0.948
 Test 0.862 0.904 0.833 0.922 0.889
  1. aPositive predictive value
  2. bNegative predictive value