Abstract: To address the issues of low real-time responsiveness and limited detection range in traditional coal roadway roof type detection, a rock type identification method based on measurement while drilling (MWD) technology was proposed in this study. Firstly, three comprehensive indexes for rock type identification, namely rock-breaking energy, logistic regression analysis, and rock hardness, were established based on drilling parameters. Secondly, a probability classification model and a random forest model were established, and the model parameters were obtained by macro-average ROC analysis. Finally, intelligent rock type identification was achieved through both the probabilistic classification model and the random forest model, which were driven by the multi-parameter comprehensive indexes. With the aid of a self-developed laboratory MWD platform, MWD experiments were conducted on spliced specimens composed of different rock types to validate the reliability of the proposed intelligent identification method. The results demonstrate that the three comprehensive indexes outperform single drilling parameters in rock type identification. The random forest model, utilizing these comprehensive indexes, accurately identifies the splicing interfaces, rock layer thicknesses, and rock types in the spliced specimens, achieving a prediction accuracy of 92.9%. Additionally, the precision, recall, and F1-score of the model all exceed 88.7%.