Data-Driven Assessment of Construction and Demolition Waste Causes and Mitigation Using Machine Learning

Construction and demolition (C&D) waste remains a critical challenge in India due to accelerated urbanisation and material-intensive construction practices. This study integrates survey-based assessment with machine learning to identify key causes of C&D waste and recommend targeted minimization strategies. Data were collected from 116 professionals representing junior, middle, and senior management, spanning age groups from 20 to 60+ years, and working across building construction, consultancy, project management, roadworks, bridges, and industrial structures. The majority of respondents (57%) had 6–20 years of experience, ensuring representation from both operational and decision-making roles. The Relative Importance Index (RII) method was applied to rank waste causes and minimization techniques based on industry perceptions. To enhance robustness, Random Forest, Gradient Boosting, and Linear Regression models were tested, with Random Forest performing best (R2 = 0.62), providing insights into the relative importance of different strategies. Findings show that human skill and quality control are most critical in reducing waste across concrete, mortar, bricks, steel, and tiles, while proper planning is key for excavated soil and quality sourcing for wood. Recommended strategies include workforce training, strict quality checks, improved planning, and prefabrication. The integration of perception-based analysis with machine learning offers a comprehensive framework for minimising C&D waste, supporting cost reduction and sustainability in construction projects. The major limitation of this study is its reliance on self-reported survey data, which may be influenced by subjectivity and regional bias. Additionally, results may not fully generalize beyond the Indian construction context due to the sample size and sectoral skew. The absence of real-time site data and limited access to integrated waste management systems also restrict predictive accuracy of the machine learning models. Nevertheless, combining industry perception with robust data-driven techniques provides a valuable framework for supporting sustainable construction management.