<p>Machine learning (ML) methods for predicting the uniaxial compressive strength (UCS) of rocks, while established, face challenges in integrating diverse geological data and dataset imbalances. This study introduces a novel Hybrid Bayesian-Group-based Machine Learning (HB-GML) method that combines Bayesian ridge regression for imputing missing data with a dynamic grouping strategy for data clustering. Utilizing a dataset of 487 samples from various lithologies, the Bayesian ridge regression was used to impute missing values, ensuring the reliability and uncertainty of the estimates were analyzed. Furthermore, the K-Nearest Neighbor-Density-based Spatial Clustering of Applications with Noise (KNN-DBSCAN) algorithm effectively clustered data into distinct groups, each reflecting its underlying data distribution and identifying noise points, without manual intervention. For each cluster, the HB-GML method identified the most optimal ML model by comparing evaluation indices across various techniques, ensuring model reliability and accuracy. The final model, integrating optimal models from all clusters, is expected to enhance predictive performance and model interpretability. Comparative analysis shows that the HB-GML method effectively clusters data based on intrinsic characteristics, offering a robust and adaptable framework for enhancing data processing, as well as significantly enhances predictive accuracy. The method's adaptive clustering strategy not only avoids data imbalances but also adapts dynamically to new data, providing a reliable solution for the practical application of ML methods in geotechnical engineering.</p>