The SDL Component Suite is an industry leading collection of components supporting scientific and engineering computing. Please visit the SDL Web site for more information....



How to use TCFEvaluator



The class TCFEvaluator simplifies the calculation of characteristic properties of binary classifiers. The general approach is as follows: first you have to load the actual data (property TrueData) and the estimated data (the classifier response property EstimData) by using the methods CopyTrueFromMatrix and CopyEstFromMatrix. Next you have to call CalcConfusionMatrix (using a specific threshold) to prepare the confusion matrix before you can use any method for obtaining various classifier performance indicators.

Optionally you can use OptimizeThresholds to determine an optimum threshold before calling CalcConfusionMatrix.

Currently, the following classifier metrics are available: Accuracy, F1Score, FnRate, FpRate, Informedness, MatthewsCorr, NegPredValue, Precision, ThreatScore, TnRate, TpRate, UncertCoeff


Last Update: 2020-Apr-13