Skyscrapers Conceptual Cost Estimation with Non-Local Data using Artificial Neural Networks and Bootstrap Aggregating

Abstract

In recent years the amount of new developed skyscraper buildings have considerable increased around the world. And due to the high construction costs of those projects, there are a lot of financial risks involved on such developments. On that scenario, accurate early cost estimates are of main importance for the successful completion of those buildings.

Preliminary research has shown that neural networks models can be great tools for accurate conceptual construction cost estimation. However, those techniques are highly dependent on considerable amount of previous similar data that can be hard to acquire in the case of skyscrapers.

With the intention to try applying the addition of non-local to estimate conceptual costs of skyscrapers, this research studied the use of neural networks in association with bootstrap aggregating (bagging) as a way to deal with the limitations on local data.

Data from 124 projects in 5 different countries was collected and different models were tested to assess the efficiency of using artificial neural networks with added non-local data and bagging. The obtained results showed promising potential for the method as a way to increase conceptual cost estimation accuracy both in countries with already sufficient data, and in countries with limited previous projects data.