Abstract: The pavement bumping was first involved in the specification "Highway Technical Condition Evaluation Standard" (JTG H20—2018) of China. However, existed research shows that the pavement bumping is also correlated with other pavement performance. This paper studied the prediction method of pavement bumping based on partial traditional pavement performances including international roughness index (IRI), rut depth (RD), pavement damage rate (DR) and pavement wear rate (WR). The pavement performance of six expressways located in Beijing was detected and evaluated, and a relationship model was built between pavement bumping and four types of pavement performance using random forest machine learning model for predicting pavement bumping sections. Results show that IRI has the greatest correlation with pavement bumping, followed by RD, and the impact of DR and WR on pavement bumping is relatively small. After cross validation and model optimization, the prediction accuracy of the pavement bumping can reach up to 99.475%, which meets the needs of practical engineering applications. This study provides a new idea for the rapid identification and detection of pavement bumping, and provides a low-cost and low resource consumption detection method of pavement bumping.