Transportation Market Rate Forecast Using Signature Transform

Currently, Amazon relies on third parties for transportation marketplace rateforecasts, despite the poor quality and lack of interpretability of theseforecasts. While transportation marketplace rates are typically verychallenging to forecast accurately, we have developed a novel signature-basedstatistical technique to address these challenges and built a predictive andadaptive model to forecast marketplace rates. This novel technique is based ontwo key properties of the signature transform. The first is its universalnonlinearity which linearizes the feature space and hence translates theforecasting problem into a linear regression analysis; the second is thesignature kernel which allows for comparing computationally efficientlysimilarities between time series data. Combined, these properties allow forefficient feature generation and more precise identification of seasonality andregime switching in the forecasting process. Preliminary result by the modelshows that this new technique leads to far superior forecast accuracy versuscommercially available industry models with better interpretability, evenduring the period of Covid-19 and with the sudden onset of the Ukraine war.