Abstract: In order to ascertain how the geological parameters will affect consumption of main materials in shield tunnelling under complex geological conditions, the random forest regression algorithm and random forest classification algorithm models have been created to conduct model training and prediction of the relation between consumption of grease, labor, water and electricity and synchronous grouting coefficient and the geological parameters. The results indicate: (1) Compressive strength is the most important geological characteristic parameter that affects consumption of labor, water, electricity and grease and is also the second most important parameter that affects the synchronous grouting coefficient; (2) Consumption of main materials in shield tunnelling is in positive correlation with compressive strength; (3) With compressive strength used as the single classification index, the quota classification standard for shield tunnelling under complex geological conditions has been established. As revealed by the classification statistics on consumption of main materials, the consumption of materials for tunnelling in soft soil stratum is approximate to the standard value, while the consumption of materials for tunnelling in hard rock is over 2 times the standard value; (4) For the non-real-time recorded data such as consumption of labor, water and electricity, the random forest regression algorithm will generate more accurate analysis results. For the real-time recorded data such as consumption of grease, the random forest classification algorithm will generate better prediction results.