This study aimed to investigate the evolution law of traveler’s mode choice behavioral decision making under new travel service model by designing behavioral experiment for travel mode decisions making under MaaS (Mobility as a Service) concept. The behavioral experiment was carried out simultaneously for two groups, with one group was provided with full information (FI) and the other group was provided with partial information (PI). By conducting behavioral characterization analysis on both individual and collective level, and constructing an improved experience-weighted attractions (EWA) learning model to analyze the learning behavioral characteristics based on the experimental data, the following findings were obtained. FI and PI had no significant impact on the evolution of the system converging to Nash equilibrium state, and the number of subjects who switched MaaS package and travel mode both steadily declined over time; there was a significant negative correlation between individual benefits and frequency of travel mode switches with PI, while there was no significant correlation between individual benefits and frequency of travel mode switches with FI; subjects with PI preferred direct response styles, whereas subjects with FI had more diverse response styles; compared to the PI group, subjects in the FI group learnt at a higher rate and were more aware of others’ adjustment their learning, and considered the unchosen strategy more often in their decisions.