This study focuses on the economic benefits of railway transportation from the aspect of data mining using convolution neural network (CNN), long short-term memory (LSTM), and multiple linear regression (MLRA) models. Improved CNN and LSTM (CNN-LSTM), and CNN and bidirectional LSTM (CNN-Bi-LSTM) models have been developed to enhance learning. The case data sets include operating mileage, and passenger and freight turnover for four transportation modes (railway, highway, aviation, and water transportation) from 1952 to 2020 in China; various evaluation indexes are used to verify model effectiveness. The CNN and LSTM model prediction error rates are 29% and 22%, respectively, verifying the role of an integrated transportation system in railway transportation and the time effect as a national economic benefit of railway transportation, respectively. The CNN-LSTM model prediction error rate is 12%, indicating that the economic benefits of railway transportation depend on the structure of various transportation modes and the time stage of transportation resource allocation. The prediction error rate of the CNN-Bi-LSTM model is 14%, suggesting the irreversibility of the impact of railway transportation resources on economic benefits. The MLRA model error rate is lower than those of the CNN and LSTM models, at 19%, but higher than those of the other models. The CNN-LSTM model is recommended to quantify the economic benefits of railway transportation. This study illustrates the systematic nature, periodicity, time lag, and irreversibility of railway transportation and economic development, providing a theoretical basis for the formulation of transportation development and investment plans.