With the deployment of China′s transportation power strategy, the high-speed railway network is continuing to expand. The demand for railway passenger travel is steadily increasing. Facing the huge high-speed railway passenger transport market, how to use intelligent and efficient deep learning technology to build a combined prediction model is a challenge. This paper integrates a variety of factors affecting ticket purchases and is committed to explore the changes in passenger ticket purchase demand on each day of the pre-sale period in real time. This shows that it′s becoming an urgent problem to provide an efficient and reliable prediction model for passenger ticket pre-sale period for the railway department. This paper takes the historical ticket purchase data of high-speed railway passengers between OD（origin-destination） on the Shanghai-Kunming high-speed railway line as an example. The combined prediction model considers the number of tickets purchased on each day of the pre-sale period of the historical continuous departure date, the date of the departure date, statutory holiday, and seasonal characteristics. The research shows that the combined prediction model based on deep learning CNN-LSTM has better predictive power than the parameter model and machine learning model, which provides a theoretical reference for the dynamic adjustment of fares in the railway passenger transport market.