RandomForest classifier validation Cross Validation Number of classes = 9 class 0.0: C1 accuracy = 0.9788 precision = 0.8746 correlation = 0.9086 errorRate = 0.0212 TruePositives = 537.0000 FalsePositives = 77.0000 TrueNegatives = 3903.0000 FalseNegatives = 19.0000 class 1.0: C2 accuracy = 0.8757 precision = 0.4940 correlation = 0.5332 errorRate = 0.1243 TruePositives = 328.0000 FalsePositives = 336.0000 TrueNegatives = 3644.0000 FalseNegatives = 228.0000 class 2.0: C3 accuracy = 0.8651 precision = 0.3947 correlation = 0.2987 errorRate = 0.1349 TruePositives = 105.0000 FalsePositives = 161.0000 TrueNegatives = 3819.0000 FalseNegatives = 451.0000 class 3.0: C4 accuracy = 0.9109 precision = 0.6297 correlation = 0.6315 errorRate = 0.0891 TruePositives = 369.0000 FalsePositives = 217.0000 TrueNegatives = 3763.0000 FalseNegatives = 187.0000 class 4.0: Nan accuracy = 0.9802 precision = 0.0000 correlation = 0.0030 errorRate = 0.0198 TruePositives = 0.0000 FalsePositives = 2.0000 TrueNegatives = 4446.0000 FalseNegatives = 88.0000 class 5.0: C5 accuracy = 0.8765 precision = 0.4953 correlation = 0.4403 errorRate = 0.1235 TruePositives = 210.0000 FalsePositives = 214.0000 TrueNegatives = 3766.0000 FalseNegatives = 346.0000 class 6.0: C6 accuracy = 0.8988 precision = 0.5655 correlation = 0.6321 errorRate = 0.1012 TruePositives = 419.0000 FalsePositives = 322.0000 TrueNegatives = 3658.0000 FalseNegatives = 137.0000 class 7.0: C7 accuracy = 0.8710 precision = 0.4772 correlation = 0.5080 errorRate = 0.1290 TruePositives = 303.0000 FalsePositives = 332.0000 TrueNegatives = 3648.0000 FalseNegatives = 253.0000 class 8.0: C8 accuracy = 0.9797 precision = 0.8841 correlation = 0.9115 errorRate = 0.0203 TruePositives = 534.0000 FalsePositives = 70.0000 TrueNegatives = 3910.0000 FalseNegatives = 22.0000 Using Testing dataset, % correct predictions = 61.8386 Total samples = 9072 RMSE = 2.2529130662212906 Bias = 0.1457231040564375 Distribution: class 0.0: C1 1112 (12.2575%) class 1.0: C2 1112 (12.2575%) class 2.0: C3 1112 (12.2575%) class 3.0: C4 1112 (12.2575%) class 4.0: Nan 176 (1.9400%) class 5.0: C5 1112 (12.2575%) class 6.0: C6 1112 (12.2575%) class 7.0: C7 1112 (12.2575%) class 8.0: C8 1112 (12.2575%) Testing feature importance score: Each feature is perturbed 3 times and the % correct predictions are averaged The importance score is the original % correct prediction - average rank 1 feature 1 : Image1 score: tp=0.5048 accuracy=0.1122 precision=0.4123 correlation=0.4803 errorRate=-0.1122 cost=-1.1611 GainRatio = 0.2871 rank 2 feature 3 : Image2 score: tp=0.4856 accuracy=0.1079 precision=0.3725 correlation=0.4729 errorRate=-0.1079 cost=-1.0772 GainRatio = 0.2756 rank 3 feature 2 : Image3 score: tp=0.3074 accuracy=0.0683 precision=0.2584 correlation=0.3299 errorRate=-0.0683 cost=-0.5009 GainRatio = 0.2215 rank 4 feature 5 : Image21 score: tp=0.1879 accuracy=0.0418 precision=0.1739 correlation=0.2134 errorRate=-0.0418 cost=-0.2770 GainRatio = 0.1302 rank 5 feature 4 : Image11 score: tp=0.1857 accuracy=0.0413 precision=0.1651 correlation=0.2157 errorRate=-0.0413 cost=-0.2557 GainRatio = 0.1567 rank 6 feature 6 : Image31 score: tp=0.1751 accuracy=0.0389 precision=0.1604 correlation=0.2028 errorRate=-0.0389 cost=-0.2515 GainRatio = 0.1115