Real-time and precise prediction of geological conditions is extremely beneficial to the safe and efficient construction of slurry pressure balanced shield machines (SPBMs), especially for super-large diameter tunnels. Machine learning algorithms provide an effective way to predict geological conditions and have been successful in hard-rock tunnel boring machine (hard-rock TBM) tunnels or earth pressure balance shield machine (EPBM) tunnels. However, the effectiveness of machine learning in predicting geological conditions for super-large diameter SPBM tunnels is still unknown. This study systematically demonstrated the application of machine learning algorithms to the prediction of geological conditions in super-large diameter SPBM tunnels based on field data from the Heyan Road crossing river tunnel. The relationship between SPBM parameters and geological conditions was first evaluated, and subsequently, an adaptive model was proposed to determine the number of input variables. The performance of three algorithms, random forest (RF), AdaBoost, and support vector machine (SVM), in rock mass classification, and seven algorithms, multi-objective random forest (MORF), single-target-RF, regressor chain-RF, single-target-AdaBoost, regressor chain-AdaBoost, single-target-SVM, and regressor chain-SVM, in formation percentage prediction, were compared. The results showed that machine learning algorithms can effectively predict the geological conditions in super-large diameter SPBM tunnels. Specifically, RF had the best performance in rock classification compared with AdaBoost and SVM. MORF outperforms the other six algorithms in terms of formation percentage prediction. Additionally, sensitivity analysis indicates that the newly proposed parameters, including slurry system parameters and cutter type, have a clear positive effect on the performance of geological condition prediction.