Individuals with alcohol use disorder (AUD) may manifest an array of neural and behavioral abnormalities, including altered brain networks, impaired neurocognitive functioning, and heightened impulsivity. Using multidomain measures, the current study aimed to identify specific features that can differentiate individuals with AUD from healthy controls (CTL), utilizing a random forests (RF) classification model. Features included fMRI-based resting-state functional connectivity (rsFC) across the reward network, neuropsychological task performance, and behavioral impulsivity scores, collected from thirty abstinent adult males with prior history of AUD and thirty CTL individuals without a history of AUD. It was found that the RF model achieved a classification accuracy of 86.67% (AUC = 93%) and identified key features of FC and impulsivity that significantly contributed to classifying AUD from CTL individuals. Impulsivity scores were the topmost predictors, followed by twelve rsFC features involving seventeen key reward regions in the brain, such as the ventral tegmental area, nucleus accumbens, anterior insula, anterior cingulate cortex, and other cortical and subcortical structures. Individuals with AUD manifested significant differences in impulsivity and alterations in functional connectivity relative to controls. Specifically, AUD showed heightened impulsivity and hypoconnectivity in nine connections across 13 regions and hyperconnectivity in three connections involving six regions. Relative to controls, visuo-spatial short-term working memory was also found to be impaired in AUD. In conclusion, specific multidomain features of brain connectivity, impulsivity, and neuropsychological performance can be used in a machine learning framework to effectively classify AUD individuals from healthy controls.