Identification the Causes of Negative Emotions in Smart Public Transportation Services Based on Social Media Data

The negative emotional comments of smart public transportation maybe contain specific reasons behind them. If the underlying reasons can be identified, it is beneficial to timely improve public transportation (PT) services and provide a more comfortable PT experience for the public. However, the application of emotional analysis based on social media in bus service evaluation is limited, especially in the exploration of the reasons behind negative evaluations of bus services, which is still in a blank stage. Therefore, based on social media data, the paper studies the method of identifying the causes of negative emotions in PT services. Sentiment data is obtained through Sina microlog. The reasons for negative emotions are proposed and classified, which are divided into seven types: long waiting times, poor hygiene, poor driver attitude, unstable driving, overcrowding, unreasonable pricing, and other reasons. Then, a classification model for negative emotions based on feature fusion and attention mechanism, BERT-BiGRU + CNN-Attention model, is established to identify the causes of various negative emotions. Finally, the proposed method is used for analysis and validation. This model incorporates an attention model on the basis of feature fusion, which assigns higher weights to important features. Therefore, the F1 value classified on the test set reached 0.9724, achieving a good identification result.