In the current study, the RF model, with a classification accuracy of 80%, identified the following multi-domain features that significantly contributed to group classification (ALC vs. CTL): 29 FC connections across key DMN nodes representing all frequency bands, three neuropsychological scores representing memory performance and all four BIS impulsivity scores. The predominant pattern of FC was hyperconnectivity (25 out of 29 connections), dominated by parahippocampal connections with other regions, indicating possible neural hyperexcitability and/or compensatory mechanism. Hypoconnectivity across prefrontal nodes (bilateral PFC and L. ACC) and other right-hemispheric nodes (IPL, LTC, and PHG) suggested possible impairments in higher-order cognitive processes. Further, the AUD group also showed poor memory performance and increased impulsivity compared to CTL individuals. The top important features from all three domains had significant associations with each other as well as with AUD status. In sum, the RF model elicited important multi-domain features that significantly contributed to the classification of ALC from CTL individuals.
