With the rapid development of China’s economy, freight volume, as a crucial indicator in transportation and supply chain management, holds significant importance in optimizing resource allocation and policymaking. To accurately forecast trends in freight volume changes, this paper proposes a deep learning model based on long short-term memory (LSTM) networks, incorporating Bayesian optimization to adjust the model’s hyperparameters, thereby enhancing forecasting performance. During the data preprocessing phase, the linear interpolation method is used to address data missing issues, ensuring the integrity and continuity of the time series data. The results demonstrate that the LSTM model based on Bayesian optimization exhibits high accuracy and stability in predicting freight volume time series. The paper provides a methodology for forecasting complex time series data and offers direction for research on data prediction problems in transportation and supply chain management concerning freight volume forecasting accuracy.