Figure 4 | Scientific Reports

Figure 4

From: Combined unsupervised-supervised machine learning for phenotyping complex diseases with its application to obstructive sleep apnea

Figure 4

The relative importance of the PSG features for comorbidity risk prediction by the RSF (n = 1754). The absolute importance for each feature was calculated through the difference between the out-of-bag prediction accuracy of the model trained on true data and the model trained on randomly permuted data for the feature. The relative feature importance was calculated by dividing the absolute feature importance of each feature by that of the feature with the highest importance, age. The top 18 features (age → central apnea) accounted for 95% of the total importance. The red mark represents the relative importance of 0.1.

Back to article page