Understanding users’ travel behaviors is an important subject in traffic science, which helps traffic management departments to formulate appropriate traffic control strategies. Travel mode identification is a critical aspect of analyzing users’ travel behaviors, aiming to precisely determine the travel modes of users and promote the sustainable development of urban intelligent transportation systems. In this article, we focus on identification methods for travel modes and propose a multi-scale convolutional neural network model based on channel attention mechanism. Our model efficiently extracts multi-scale sub-features and prioritizes key features. Additionally, to deeply extract effective information from the GPS data, we design 24 time-domain features and frequency-domain features and simplify them using principal component analysis and random forest algorithm. To verify the effectiveness of our method, we design five traditional machine learning models and two shallow deep learning models as comparative models, which are trained and tested on the Geolife dataset. Experimental results show that our approach is superior to the baseline models, achieving a maximum accuracy of 83.09%.