Reservoir characterization remains a major challenge for Quantitative Interpretation, particularly in complex geological settings. We have developed a comprehensive Machine Learning (ML) methodology for lithology and fluid identification that consists of two parts. Part one deals with geophysics-based data preparation and augmentation, where rock physics information is utilized to simulate the seismic responses of different reservoir rock and fluid properties. The second part involves optimizing numerous ML algorithms and feature sets to obtain the best prediction scores. Case studies from four different fields in Malay Basin covering different siliciclastic depositional environments are provided. Well logs and seismic inversion predictions prove that up to 80 accuracy on blind well tests are achievable. The results also help highlight regions of high or low prediction accuracy. Potential applications for this method include prospect de-risking as well as near field exploration. © 81st EAGE Conference and Exhibition 2019. All rights reserved.