<p>During Baijiu brewing, the physicochemical parameters of Daqu serve as crucial indicators for evaluating its quality. However, traditional testing methods require sample destruction and lack real-time monitoring capabilities, making it difficult to meet the demands of intelligent production in the modern brewing industry. This study was aimed at proposing a rapid, accurate, and nondestructive testing method for Daqu physicochemical parameters based on machine learning models. The environmental parameters and physicochemical data of Daqu were collected to extract the key environmental time period features using XGBoost, RF, and ReliefF algorithms. Also, a support vector regression (SVR) model was constructed to predict acidity and moisture content. The prediction accuracy of acidity was further enhanced by introducing a regression chain (RC) optimization algorithm combined with a backpropagation neural network (BPNN) to establish a weighted integrated prediction model. The XGBoost–SVR model demonstrated the best predictive performance for moisture content and acidity, with <i>R</i>² values reaching 0.9686 and 0.8776, respectively. The optimization through the RC–BPNN integrated model improved the <i>R</i>² for acidity prediction by 4.5%. This method provides effective technical support for the real-time, nondestructive monitoring of the Daqu fermentation process, thereby contributing to the intelligent development of Baijiu brewing.</p>