Ecodrive-Deep Learning Models For Accurate Prediction of Vehicle Co2 Emissions

This research study presents a novel approach utilizing advanced deep learning models for accurate prediction of vehicle CO
2
emissions, addressing a critical challenge in environmental management and policy-making. Conventional methods often struggle to capture the intricate dynamics of emissions accurately. To overcome this limitation, we propose the utilization of CNN-LSTM and DCNN architectures. Training and testing of these models were conducted using an extensive dataset encompassing various factors influencing CO
2
emissions, including vehicle characteristics and driving conditions. The empirical findings validate that the recommended models substantially enhance the accuracy of predictions. Specifically, DCNN outperformed KNN significantly, with a test accuracy of 96.34% compared to KNN’s 74.02%. Similarly, DCNN exhibited higher accuracy metrics in both test and validation sets compared to LSTM, with a test accuracy of 96.21% versus 93.72% and a validation accuracy of 97.71% versus 94.89%, respectively. Furthermore, DCNN showcased higher accuracy metrics compared to RF, with a test accuracy of 96.34% versus 92.55% and a validation accuracy of 97.76% versus 95.49%. DCNN also outperformed RNN in both test accuracy (95.39% vs. 91.10%) and validation accuracy (97.66% vs. 94.59%), with a final model accuracy significantly favoring DCNN. The findings of this study hold considerable importance for directing sustainable measures and shaping policies to lessen the effects of climate change.