Comparing two AI methods for predicting the future trend of New Zealand building projects: Decision Tree and Artificial Neural Network

Abstract

The rise of Artificial Intelligence and Machine Learning in many aspects of construction management has helped this industry to further improve the management, design, and planning of construction projects. This trend happens in many construction sectors, including in New Zealand. Whilst relatively smaller compared to construction sectors in other OECD countries, the construction sector in New Zealand carries a similar degree of complexity and with its own unique characteristics. Various studies showed that AI and ML can be used for analysis of construction data to generate further insights and to predict future trends in construction sectors. However, the AI approaches have their own set of challenges such as complexity, high cost of training, failure, and change. Aiming to better understand the trends and requirements of New Zealand building projects, this study started with a review of the existing AI methods that are currently being applied. Accordingly, compare and evaluate the accuracy of two AI prediction methods. The two methods of Decision Tree and Artificial Neural Network are selected based on their predictive power and accuracy. These methods are conducted by using available historical building data which is available on StatsNZ website. A portion of the data is used for testing and evaluation purposes, and the rest of the data is used for training the AI methods. It was identified that the Decision Tree method did not show suitable accuracy for prediction building consents issued data. In comparison, Artificial Neural Network shows a reasonable range with 95% of confidence level. Therefore, this method is applied for building consents issued in New Zealand.