Fig. 4. Random Forest Model using data from highly integrated sensing chip.

a Confusion matrix of the ion-type classifier. b Top four SHAP (Shapley Additive exPlanations) values for the ion-type classifier. $V_{\mathrm{DP}}^{\mathrm{Na}}$, $V_{\mathrm{DP}}^{\mathrm{Ca}}$ and $V_{\mathrm{DP}}^K$ represent the Dirac Points of devices with Na⁺ ISM, Ca²⁺ ISM, and K⁺ ISM functionalization respectively. $g_{m,\max }^K$ represents the maximum slope of devices with K⁺ ISM functionalization. c–e Confusion matrix of the multi-output classification model for K⁺, Na⁺, and Ca²⁺ ion concentration in multi-ion solutions. f I–V characteristics of graphene sensor array tested in water (yellow curves) and in human serum (blue curves). g Impact of chip usage on classification accuracies for the pure solutions test.