A computer-implemented method for early prediction of antimicrobial resistance (AMR) emergence using machine learning applied to longitudinal antibiogram time series. A gradient boosting machine (GBM) and long short-term memory (LSTM) neural network are trained on historical susceptibility records to generate early-warning signals before resistance crosses clinical thresholds. Features include resistance rate trends, volatility, and cross-drug correlations. The system outputs facility-level AMR risk scores.