Several factors (e.g., interpersonal stress, affect) prospectively predict loss-of-control eating (LOCE) and overeating in adolescents, but most past research has tested predictors separately. We applied machine learning to simultaneously evaluate multiple possible predictors of future LOCE and overeating in pooled and person-specific models. Twenty-six adolescents (80.77% female; age = 15.86 1.65 years, BMI %ile = 92.77 9.05) who endorsed 2 past-month LOCE episodes completed EMA ratings over one week. Pooled models were fit to the aggregated data with elastic net regularized regression and evaluated using nested cross-validation. Person-specific models were fit and evaluated as proof-of-concept. Across adolescents, the median out-of-sample R2 of the pooled prediction models to identify LOCE was 0.21. The top predictors were between-subjects craving, feeling sad, being in a conflict, stress level (inverse association), feeling angry/mad (inverse association), and within-subjects wishing relationships were better. The median out-of-sample R2 for the pooled overeating model was 0.28. The top predictors were between-person craving, feeling lonely, mixed race, and feeling rejected (inverse association), and within-subjects guilt, nervousness, and feeling scared (inverse association). Person-specific models demonstrated poor fit (median LOCE R2 = .002, median overeating R2 = -.002); 57% and 40% of adolescents’ models performed better than chance for LOCE and overeating, respectively. Altogether, group-level models hold utility in predicting future LOCE and overeating, but model performance for person-specific models is variable. Ultimately, a mix of these approaches may improve the identification of momentary predictors of LOCE and overeating, providing novel and personalized opportunities for intervention.