Determine the winning classifiers after training per dataset/resolution

[elboyran]

Add a new Excel sheet to the C:\Projects\DynaSlum\Results\Classification3Classes\PerformanceComparision\ClassifiersPerformance.odt

[elboyran]

Use a single score like the F-score? F = 2*(precission*recall/(precission+recall))

[elboyran]

The F score is added to the Excel book but highest Recall + high Precision are used to determine winning classifier

[elboyran]

 Dataset #   Tile size  Best classifier Training           Precission   Recall  F score
1   417 pixels =  250 m.    1.21 Ensemble/ Subspace KNN        92.0%    100.0%  0.9583
2   333 pixels =  200 m.    1.12 KNN/ Fine KNN ?                   92.0%    78.0%   0.844 (not the best)
3   250 pixels =  150 m.    1.12 KNN/ Fine KNN                 93.0%    87.0%   0.899
4   167 pixels = 100 m. 1.12 KNN/ Fine KNN                 78.0%    63.0%   0.697