Events News Research CBS CBS Publications Bioinformatics
Staff Contact About Internal CBS CBS Other

Calculating performance of a prediction method




Always estimate performance on an independent test set of sequences, which is non-similar to the training set.

Calculate:
 

  • True Positives (TP)      positive in reality, positive in prediction
  • True Negatives (TN)     negative in reality, negative in prediction
  • False Positives (FP)      falsely predicted to be positive (negative in reality)
  • False Negatives (FN)     falsely missed by prediction (positive in reality)


Various measures of performance:

  • Sensitivity  tp/(tp + fn)
  • Specificity  tn/(fp + tn)
  • Correlation coefficient