This article aims to design a subway investment management platform based on DL (Deep Learning) to improve subway projects' investment efficiency and risk management levels. The article adopted the DL algorithm to comprehensively analyze historical subway investment data and constructed prediction and risk assessment models to achieve this goal. The application effect of the DL algorithm in subway investment management and the practicality of the platform were verified through simulation experiments. This article first collected a large amount of historical subway investment data and processed and learned these data using the DL algorithm. By constructing an NN (Neural Network) model, accurate prediction of investment return rate and market demand for subway projects has been successfully achieved. Furthermore, a risk assessment model was also constructed to quantify and evaluate potential risks in subway projects. The experimental results indicate that the DL algorithm has significant prediction and risk assessment advantages. Compared with traditional statistical learning methods and machine learning algorithms, the DL algorithm can more accurately predict investment return and market demand, providing users with more reliable decision support. In addition, the risk assessment model effectively helps users understand the risk status of the project and formulate corresponding risk management measures.