Modeling the trend of construction materials industry with NARNETs

Abstract

The price of materials is dependent on different factors such as raw material costs, production costs, and cost of logistics. The construction industry professionals face difficulties in times when there are fluctuations in material prices. This study aimed to model the expectancy of trends in material prices through a time series analysis, as the expectancy of trend is a time dependent dataset. . In this context, the study is focused on utilization of a special type of ANN (and special type of RNN) architecture known as Nonlinear Autoregressive Neural Network (NARNET). Ten different NARNET configurations were implemented in MATLAB and their performance were tested in the study. The results have shown that NARNETs are able to model the expectancy of trend accurately.