The performance of the brake system is directly related to the safety and reliability of the mine hoist operation. Mining the useful fault information in the operation of a mine hoist brake system, analyzing the abnormal parts and causes of the equipment, and making accurate early prediction and diagnosis of hidden faults are of great significance to ensure the safe and stable operation of a mine hoist. This study presents a fault diagnosis method for hoist disc brake system based on machine learning. First, the monitoring system collects the information of the hoist brake system, extracts the fault features, and pretreats it by SPSS (Statistical Product and Service Solutions). This work provides data support for fault classification. Then, due to the complex structure of the hoist brake system, the relationship between the fault factors often has a significant impact on the fault. Considering the correlation between the fault samples and the attributes of each sample data, the C4.5 decision tree algorithm is improved by adding Kendall concordance coefficient, and the improved algorithm is used to train the sample data to get the decision tree classification model. Finally, the fault sample of the hoist brake system is trained to get the algorithm model, and then the fault diagnosis rules are generated. The state of the brake system is judged by classifying the data. Experiments show that the improved C4.5 decision tree algorithm takes the relativity of conditional attributes into account, has a higher diagnostic accuracy when processing more data, and has concise and clear fault classification rules, which can meet the needs of hoist fault diagnosis.