Measure While Drilling (MWD) data, a high-resolution sensor dataset collected during rock tunnel excavation worldwide, is underutilised, mainly serving for geological visualisation. Recent studies have demonstrated that MWD-data can identify rock type and rock mass quality for data from short blasting holes. Currently, there are no reliable, efficient methods in tunnelling to accurately predict rock mass characteristics in advance, which limits planning of advance rock support, excavation design, and the logistical planning for reuse of excavated rock. Traditional exploratory methods, such as geophysics, core drilling or subjectively interpreted hammer holes, suffer from being time-consuming, costly, and offering insufficient resolution. To overcome these limitations, we employed ensemble machine learning models on MWD sensor data from 24 m exploratory holes in long infrastructure tunnels with diverse geology. This approach enables accurate predictions of rock type and rock mass quality with a 1-meter resolution, providing a planning horizon of several days. Additionally, we introduce confidence metrics for each meter of prediction, enhancing decision support by identifying areas of uncertainty. Our approach, particularly focusing on the challenging geological transition zones, achieved balanced accuracies above 0.8 for rock quality and 0.9 for rock type. This advancement offers a significant improvement in the planning and execution of rock tunnelling projects.