Analysis of the Impact of Weather on Dynamic Ordering Through XAI

The objectives of this study are to optimize delivery strategies and estimate CO2 emission reductions. The results of the analysis showed, first, that the Gradient Boosting model is optimal for predicting same-day demand incorporating weather data. Second, it was quantitatively shown that the choice of delivery method based on the clustering results contributes to environmental impact reduction. The novelty of this study is the use of XAI to improve the transparency of the model and incorporate external environmental factors into the delivery planning. It is hoped that the selection of dynamic modeling and delivery methods will lead to a better society.