In the subjective evaluation of noise, it is often necessary for experienced people to listen to a certain number of sound samples and score and evaluate it, which is time-consuming and labor-intensive, and it is easy to cause hearing fatigue and even lead to inaccurate evaluation results. In the past research on sound quality prediction models, multiple linear regression models, support vector machine models, neural network models, etc. were mostly used for prediction. In this paper, psychoacoustic objective parameters such as loudness, sharpness, roughness, speech intelligibility, and A-weighted sound pressure level are selected for objective evaluation and used as the input value of the prediction model. A grade scoring method was used to evaluate 68 samples of vehicle interior noise. Finally, the arithmetic average of the subjective scores of 20 people was calculated and used as the output value of the subsequent prediction model. Based on the XGBoost algorithm, a prediction model