Forecasting daily Platts price of auto diesel using time series and neural network approaches

Abstract

Predicting price fluctuations of petroleum products are challenging as it is irregular, non-linear and complex. However, accurate predictions are vital to minimize the economic loss. The objective of this research work is to identify the most appropriate model out of time series and deep learning neural network models to forecast daily Platts price per barrel of auto Diesel. The daily Platts prices from January 2010 to March 2021 were collected from the Ceylon Petroleum Corporation. Initially, price movements were observed using descriptive statistics.