Figure 8

The predictive performance of machine learning and normalization methods across data sets. The heatmap visualize the number of data sets in which one pair of methods (row) performed significantly better than another pair of methods (column). The Wilcoxon signed-rank test was used to compare the error rates of all combinations containing one pair against the error rates of all combinations containing the other pair. Significance was determined using a p-value threshold of 0.05.