Skip to main content
Fig. 5 | BMC Medical Informatics and Decision Making

Fig. 5

From: Seasonally adjusted laboratory reference intervals to improve the performance of machine learning models for classification of cardiovascular diseases

Fig. 5

(a-c) Plot of SHAP values as estimate of feature contribution by ICD-10 chapter IX code for the versions 1 and version 2 Neural Net model for the three most important features across all of the ICD-10 codes. ICD-10 codes are sorted by absolute mean SHAP value, i.e., from left to right the contribution of the laboratory test increases. Shaded bands indicate there was no significant contribution difference for the given level 3 ICD-10 code/ model version (p value < 0.05, Bonferroni corrected). Supplementary Table 5 reports all mean absolute SHAP values for Neural Net version 1 and 2 models

Back to article page

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

Contact us