Figure 9
From: A p-Median approach for predicting drug response in tumour cells
Comparison of confidence intervals on prediction accuracy. Results for Sherf and Liu dataset are reported (red and blue series respectively). Accuracy of traditional classifiers, i.e. DT, 1-NN, NB and SVM, are based on train/infer one drug at a time by using all the available genes. Results corresponding to the clustering approaches are concerned with BNs trained according to the clustering output and the IG feature selection policy.