Deconstructability prediction for building using machine learning and ensemble feature selection techniques

TL;DR

This paper introduces a machine learning model to predict whether buildings can be deconstructed for reuse, aiming to support circular economy practices.

Contribution

The novelty lies in developing a predictive model using machine learning and ensemble feature selection for assessing building deconstructability.

Findings

A predictive model was developed to assess building deconstructability.

The model uses machine learning and ensemble feature selection techniques.

The model's application was demonstrated through a real-world deconstruction project.

Abstract

Construction industries remain one of the most significant users of materials and generators of waste in the UK and globally. Notwithstanding, the principle of circular economy is becoming prominent as an effective means for powering greater resource efficiency. It has the prospect of unlocking significant economic value, particularly at the building end of useful life through reuse. A noteworthy end-of-life practice which aligns with this idea is deconstruction, which is the careful disassembly of the building into components and sub-components for reuse. However, deconstruction is not meant for all buildings, and this is because a typical building is constructed as a permanent product waiting to be disposed of after use. Laying on this foundation, assessing the building for deconstruction is necessary, and it is mainly done via several manual inspections, which may be expensive and time-consuming. A deconstructability predictive model using a machine learning-based model and ensemble feature selection techniques was developed to tackle this problem. This paper elaborates on the model creation and illustrates its application through a real-world deconstruction project.

The online version contains supplementary material available at 10.1038/s41598-025-00790-0.

Figures