Neural Networks as an Alternative Tool for Predicting Fossil Fuel Dependency and GHG Production in Transport

Abstract

This paper deals with problems of the comparative analysis of results provided by the processing of predicting trends in freight transport, especially concerning dependency on fossil fuels and GHG (greenhouse gas) production. This topic has been in compliance with current requests for sustainable transport and the environment both being strongly emphasized in recent EU directions. Based on publicly available statistical data covering the selected time period, two completely different predicting methods—neural networks and mathematical statistics—are used for forecasting both of the above-mentioned trends. Obtained results are further analyzed from viewpoints of value concordance and reliability. The concluding comparative analysis summarizes the pros and cons of both approaches. Given the fact that forecasting methods generate model values representing future events whose states are unverifiable, the presented procedure can be used as a tool for verifying of their verisimilitude by means of comparing results obtained by the above-mentioned methods.