/// <summary>
/// Scores a set of predictions against the actual values.
/// </summary>
/// <param name="predictions">Predicted values.</param>
/// <param name="actual">Actual values.</param>
/// <param name="truthLabel">(Optional) the truth label in the <paramref name="actual"/> vector.</param>
/// <param name="falseLabel">(Optional) the false label in the <paramref name="actual"/> vector.</param>
/// <returns></returns>
public static Score ScorePredictions(Vector predictions, Vector actual,
double truthLabel = Ject.DefaultTruthValue, double falseLabel = Ject.DefaultFalseValue)
{
var score = new numl.Supervised.Score()
{
TotalPositives = actual.Where(w => w == truthLabel).Count(),
TotalNegatives = actual.Where(w => (w == falseLabel || w != truthLabel)).Count(),
TruePositives = actual.Where((i, idx) => i == truthLabel && i == predictions[idx]).Count(),
FalsePositives = actual.Where((i, idx) => (i == falseLabel || i != truthLabel) && predictions[idx] == truthLabel).Count(),
TrueNegatives = actual.Where((i, idx) => (i == falseLabel || i != truthLabel) && predictions[idx] != truthLabel).Count(),
FalseNegatives = actual.Where((i, idx) => i == truthLabel && (predictions[idx] == falseLabel || predictions[idx] != truthLabel)).Count(),
Examples = predictions.Length
};
score._IsBinary = actual.IsBinary();
// if the labels are continuous values then calculate accuracy manually
if (!score._IsBinary)
{
score._totalAccuracy = (predictions.Where((d, idx) => d == actual[idx]).Count() / predictions.Length);
}
score.RMSE = Score.ComputeRMSE(predictions, actual);
score.CoefRMSE = Score.ComputeCoefRMSE(predictions, actual);
score.NormRMSE = Score.ComputeRMSE(predictions, actual);
score.MeanAbsError = Score.ComputeMeanError(predictions, actual);
score.SSE = Score.ComputeSSE(predictions, actual);
score.MSE = Score.ComputeMSE(predictions, actual);
score.CrossEntropy = Score.ComputeCrossEntropy(predictions, actual);
return(score);
}
/// <summary>
/// Scores a set of predictions against the actual values.
/// </summary>
/// <param name="predictions">Predicted values.</param>
/// <param name="actual">Actual values.</param>
/// <param name="truthLabel">(Optional) the truth label in the <paramref name="actual"/> vector.</param>
/// <param name="falseLabel">(Optional) the false label in the <paramref name="actual"/> vector.</param>
/// <returns></returns>
public static Score ScorePredictions(Vector predictions, Vector actual,
double truthLabel = Ject.DefaultTruthValue, double falseLabel = Ject.DefaultFalseValue)
{
var score = new numl.Supervised.Score()
{
TotalPositives = actual.Where(w => w == truthLabel).Count(),
TotalNegatives = actual.Where(w => (w == falseLabel || w != truthLabel)).Count(),
TruePositives = actual.Where((i, idx) => i == truthLabel && i == predictions[idx]).Count(),
FalsePositives = actual.Where((i, idx) => (i == falseLabel || i != truthLabel) && predictions[idx] == truthLabel).Count(),
TrueNegatives = actual.Where((i, idx) => (i == falseLabel || i != truthLabel) && predictions[idx] != truthLabel).Count(),
FalseNegatives = actual.Where((i, idx) => i == truthLabel && (predictions[idx] == falseLabel || predictions[idx] != truthLabel)).Count(),
Examples = predictions.Length
};
score._IsBinary = actual.IsBinary();
// if the labels are continuous values then calculate accuracy manually
if (!score._IsBinary)
{
score._totalAccuracy = (predictions.Where((d, idx) => d == actual[idx]).Count() / predictions.Length);
}
score.RMSE = Score.ComputeRMSE(predictions, actual);
score.CoefRMSE = Score.ComputeCoefRMSE(predictions, actual);
score.NormRMSE = Score.ComputeRMSE(predictions, actual);
score.MeanAbsError = Score.ComputeMeanError(predictions, actual);
score.SSE = Score.ComputeSSE(predictions, actual);
score.MSE = Score.ComputeMSE(predictions, actual);
return score;
}
