Fig. 2: Feature importance plots for substance use and SUD behavior prediction analyses. | npj Mental Health Research

Fig. 2: Feature importance plots for substance use and SUD behavior prediction analyses.

From: Predicting substance use behaviors with machine learning using small sets of judgment and contextual variables

Fig. 2

Balanced Random Forest (BRF) analyses were used to classify recent versus not-recent classes for all substance use and SUD behavior variables. The predictors are arranged according to the mean decrease in Gini scores with the most important predictors for classification listed on the top. Plots correspond to each of the eight questions (as listed in Table 2) (a) alcohol, (b) cannabis, (c) opioid, (d) stimulant, (e) time spent, (f) social problems, (g) isolation and (h) withdrawal symptoms.

Back to article page