Abstract: The Jining area of Inner Mongolia is one of the significant molybdenum polymetallic mining regions along the northern margin of the North China Craton, exhibiting considerable potential for molybdenum polymetallic mineralization. However, the Cenozoic basalts and clastic sediments covering the surface in this area obscure and attenuate mineralization-related information. Traditional lithology identification methods relying solely on a single type of data struggle to meet the complex demands of resource exploration in overburden regions. Therefore, this study integrated remote sensing and gravity-magnetic anomaly data to conduct lithology identification research using a random forest algorithm optimized iteratively via Bayesian optimization for the Jining shallow overburden area. Remote sensing and gravity-magnetic anomaly data from the study area were processed and fused with other multi-source data. Then, a test area was established to analyze the effectiveness of the intelligent lithology identification method based on multi-source data. Through accuracy evaluation, the classification performance of different feature combinations was compared and analyzed, and the importance of various features in lithology classification was assessed. Finally, the Bayesian-optimized random forest algorithm was applied to lithology classification using fused multi-source data across the entire study area. The results indicate that compared to using only a single type of data, this method yields the optimal lithology identification outcome, with F1 value reaching 0.885. It also refines the distribution of granite in the study area, resulting in clearer lithological boundaries. This research provides valuable insights and guidance for geological exploration in the region.