Evaluation system for urban traffic intelligence based on travel experiences: A sentiment analysis approach

Precise and comprehensive evaluation of urban traffic intelligence plays a vital role in the development of intelligent transportation systems. However, the majority of existing evaluation methods primarily rely on physical measurements, thereby overlooking the travel experiences of traffic participants. This results in a significant discrepancy between the expected outcomes of transportation design and the actual perceived travel experiences. Therefore, this study proposes a data-driven evaluation system for urban traffic intelligence based on travel experiences. In particular, the travel experiences of the public are extracted from social media data and evaluated by a sentiment analysis approach. Firstly, an indicator library is established through literature research, and it is further enhanced by a survey to ensure its comprehensiveness. After that, the text data scraped from social media posts is classified into the corresponding indicators via a pre-trained language model. We then employ a lexicon-based model to conduct sentiment analysis on the classified text data. Specifically, the lexicon-based model can not only identify the polarity of the text data but also determine the intensity of the sentiment expressed. To address the imbalanced distribution of social media data, we employ the oversampling technique to correct the data skewness. The proposed method is tested in Shanghai, China, and the results demonstrate consistency with those obtained from the analytic hierarchy process with survey data. Furthermore, the sentiment analysis approach exhibits stable performance even when provided with a limited amount of input data. The evaluation results indicate that the information accessibility and flexibility of urban transportation in Shanghai are satisfactory. However, there is a need for further improvement in the areas of safety, comfort, and affordability based on the analysis of travel experiences.