Quantification analysis and prediction model for residential building construction waste using machine learning technique

Prediction of construction waste is one of the successful techniques to reduce the amount of waste generation at source. Estimation of construction waste at each stage or phase of project is very essential to accurately compute and predict the total waste generation. The study aims to quantify the amount of construction waste at different stages of construction project so as to develop a machine learning model to accurately predict the amount of generated waste at various stages and from variable sources. About 134 construction sites were inspected to collect the generated waste data. As the construction activities are very dynamic in nature, it is very important to precisely compute the waste generation, to analyze the data prediction model, and enable to predict the sources and the amount of waste likely to generate. The decision tree and the K-nearest neighbors algorithm are used for analyzing, and the neural networks performance was studied by providing gross floor area and material estimation. The results indicate that an appreciable amount of waste is generated at every stage of project having considerable high cost, and a particular pattern has been observed for waste materials at typical stages of projects. The model has average RSME values of 0.49 which indicates the accuracy of model is satisfactory for use to perform the predictions. The combined average accuracy of the decision tree and KNN was found to be 88.32 and 88.51, respectively. These findings can provide basic data support and reference for the management and utilization of construction waste. © 2023, The Author(s), under exclusive licence to Springer Nature Switzerland AG.