Diesel Price Prediction Models with Traditional Time-Series Algorithms and Neural Networks

Diesel fuel price is generally considered to be one of the core factors that contributes to the cost of transportation and logistics activities. However, stand-alone time-series algorithms without considering other related logistics attributes may not be sufficient to construct well-defined diesel priceprediction models. This work used artificial Deep Neural Networks (DNNs) to construct prediction models with consideration of the Road Freight Transport Index (RFTI) and the Baltic Dry Index (BDI) as model attributes. The results from regression models with BDI and RFTI attributes were compared against models employing traditional time series algorithms, such as moving average, Autoregressive Integrated Moving Average (ARIMA), exponential smoothing, and seasonal forecasting. Proposed diesel price prediction models with consideration of BDI and RFTI can perform better than traditional time series in the end.