Figure 4 | Scientific Reports

Figure 4

From: Highly accurate machine learning prediction of crystal point groups for ternary materials from chemical formula

Figure 4

The obtained classification measures for the considered 49 sole and hybrid classification schemes. The main columns from left to right correspond to BA, UM, and \(1-\left| \text {B}-1 \right| \). On the other hand, the rows correspond to the weighted cross-validated mean, best, and worst models. The point groups are divided into two sets depending on whether they have a high positive count or low positive count. The threshold is 19K positive counts, which represents 1% of the total number data. The vertical (horizontal) axis refers to the classifier applied on the high (low) positive count set. For each panel, the diagonal cells illustrate the scores using a single type of classifier for both sets while the off-diagonal cells consist of the scores from a distinct pair of classifiers.

Back to article page