Analysing and quantitative examination for development of predictive frameworks in residential construction waste by using machine learning models

Abstract

This article centres on the reduction of construction waste through the identification of its sources, accurate waste measurement at project phases, and accurate prediction of waste generation throughout the construction process. Emphasis is placed on the significance of source identification and waste estimation at each project stage to precisely calculate overall waste. The article identifies and categorizes key factors contributing to waste generation, employing the Relative Importance Index (RII) method to determine their significance, severity, and contribution to waste generation. The article delves into the findings to uncover key contributors to trash development across the different phases of construction. These results provide important information for planning waste reduction initiatives. Furthermore, the study delves into the use of an estimating method to quantify the waste generated by key civil engineering materials throughout three distinct phases of a project. Results from this quantification reveal that at the substructure stage sand and bricks, at the superstructure stage bricks, and at the finishing stage external wall finishes experience the highest quantities of waste. Leveraging data from 134 construction sites, the research creates a machine learning model to precisely anticipate waste. The K-NEAREST NEIGHBOR algorithm has an average RMSE of 0.36 and the decision tree method has an average RMSE of 0.41. The model's 88% accuracy supports construction waste management and use. This research uses machine learning and data analysis to quantify and anticipate building waste at various project phases. The study's features and model accuracy enhance construction waste management techniques and provide significant insights for minimising waste throughout the building life cycle.