Using Univariate Models for Construction Output Forecasting: Comparing Artificial Intelligence and Econometric Techniques

Abstract

Forecasting of the volume of construction works plays an important role in ensuring that stakeholders (contractors, consultants, government, etc.) formulate policies for strategic long-term planning in the construction sector. In construction economics, research efforts targeted at construction output forecasting have been limited due to the lack of availability of quantitative data. To address this problem, the current study explores the use of univariate modeling techniques in construction output forecasting. Three univariate modeling techniques [namely, Box-Jenkins, neural network autoregression (NNAR), and support vector machine (SVM)] were used to predict output of various construction sectors. In the literature, the Box-Jenkins model is considered a benchmark univariate method due to its simplicity, sound theoretical basis, and predictive performance. Out-of-sample forecasting was used to evaluate the predictive accuracy of the developed models. The results of the study revealed that the SVM model can reliably and accurately predict residential, maintenance, and total construction output in the medium term. The univariate models reported here can be implemented using historical data on construction output in other countries; this is particularly useful in cases where data are inadequate for multivariate models. The developed SVM model can serve as a tool for estimating future trends in construction output.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 25 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.