Rock abrasivity is one of the crucial factors of TBM cutter wear. On the basis of previous studies, this paper systematically analyzes and summarizes the influencing factors of rock Cerchar abrasivity index (CAI), which can be divided into two categories: rock geo-mechanical parameters and experimental environmental factors. The former includes rock equivalent quartz content (EQC), rock strength, hardness, water content and P-wave velocity, while the latter mainly includes surrounding stress, steel stylus hardness, rock surface conditions and scratch length. In the geo-mechanical parameters of rocks, the cerchar abrasivity index rises with the increase of EQC, strength and hardness of rocks; the CAI value of rocks saturated with water is lower than that of dry rocks; P-wave velocity can be seen as a rough reflection on the rock abrasivity, which shows a good correlation with it. Then, the feature importance weight of the two decisive factors, i. e., EQC and strength of rocks, was analyzed by XGBoost machine learning method, and the two weight values were 0.723 and 0.277, respectively. The XGBoost analysis results show that the prediction model has a good prediction effect.