Modeling tailpipe emissions presents complex challenges due to its inherent multidimensional nature and complex time-dependent behavior between variables. In this thesis, five machine learning models applied to multivariate time series prediction are comparatively analyzed to determine their effectiveness in predicting tailpipe emissions. These models include LSTM, Seq2seq with LSTM, CNN-LSTM, PredGAN, and Spacetimeformer. experimental results show that the CNN-LSTM model performs well in this area. In addition, this study reveals the potential factors that lead to model discrepancies and concludes with some recommendations to enhance the application of machine learning in future research for more robust tailpipe emission modeling.
