Neural Networks Model for Prediction of Construction Material Prices in Egypt Using Macroeconomic Indicators

Abstract

Adequate cost estimation at the planning phase is an integral part of a construction project’s success. Many uncertainties disturb the planners’ initial estimations and lead to cost overruns. Although many researchers highlighted the correlation between economic conditions and construction costs, accurate quantification of the impact of this correlation has not yet been reached. This paper proposes three models that utilize artificial neural networks (ANNs) to predict the future prices of major construction materials, namely steel reinforcement bars and portland cement, in the context of the Egyptian construction industry 6 months ahead. A Microsoft Excel spreadsheet that also utilizes genetic algorithm (GA), NeuralTools software, and Python programing language in Spyder software was used to develop the three models. Historical data of steel and cement prices as well as macroeconomic indicators in Egypt from May 2008 to June 2018 were used for training, testing, and validation of the proposed models. The inputs to the proposed ANN models are the identified leading economic indicators such as gross domestic product, unemployment rate, and Consumer Price Index (CPI). The developed ANN models show promising results in prediction of month-to-month variations in material prices while having mean-absolute-percentage error that ranges from 4.0% to 11% for the different models. The proposed models can potentially be useful tools for construction contractors as well as owners for predicting and quantifying the fluctuations of major construction materials prices to prepare mitigation measures that will reduce the extra costs incurred.