Any idea why SVM isn’t separating classes? I’ve also carried out the same analysis using NBC and it gives me the results that I would expect.
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Predicted list size : 3315
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: ValidationLabeledListSample size : 3315
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Training performances:
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Confusion matrix (rows = reference labels, columns = produced labels):
[1] [2] [3] [4] [5]
[ 1] 1020 0 0 8 0
[ 2] 82 0 0 565 0
[ 3] 138 0 0 24 0
[ 4] 49 0 0 1236 0
[ 5] 0 0 0 193 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Precision of class [1] vs all: 0.791311
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Recall of class [1] vs all: 0.992218
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: F-score of class [1] vs all: 0.880449
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Precision of class [2] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Recall of class [2] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: F-score of class [2] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Precision of class [3] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Recall of class [3] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: F-score of class [3] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Precision of class [4] vs all: 0.610069
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Recall of class [4] vs all: 0.961868
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: F-score of class [4] vs all: 0.746602
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Precision of class [5] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Recall of class [5] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: F-score of class [5] vs all: 0
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Global performance, Kappa index: 0.502801
2023-01-10 11:42:09 (INFO) TrainVectorClassifier: Execution took 1.106 sec
This is for NBC
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Predicted list size : 3315
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: ValidationLabeledListSample size : 3315
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Training performances:
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Confusion matrix (rows = reference labels, columns = produced labels):
[1] [2] [3] [4] [5]
[ 1] 1019 8 0 1 0
[ 2] 31 519 7 90 0
[ 3] 0 0 162 0 0
[ 4] 0 202 34 1049 0
[ 5] 0 0 0 0 193
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Precision of class [1] vs all: 0.970476
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Recall of class [1] vs all: 0.991245
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: F-score of class [1] vs all: 0.980751
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Precision of class [2] vs all: 0.711934
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Recall of class [2] vs all: 0.802164
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: F-score of class [2] vs all: 0.75436
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Precision of class [3] vs all: 0.79803
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Recall of class [3] vs all: 1
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: F-score of class [3] vs all: 0.887671
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Precision of class [4] vs all: 0.920175
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Recall of class [4] vs all: 0.816342
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: F-score of class [4] vs all: 0.865155
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Precision of class [5] vs all: 1
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Recall of class [5] vs all: 1
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: F-score of class [5] vs all: 1
2023-01-10 11:41:40 (INFO) TrainVectorClassifier: Global performance, Kappa index: 0.843544