Estimation of construction waste composition based on bulk density: A big data-probability (BD-P) model

Estimating the composition of construction waste is crucial to the efficient operation of various waste management facilities, such as landfills, public fills, and sorting plants. However, this estimating task is often challenged by the desire of quickness and accuracy in real-life scenarios. By harnessing a valuable data set in Hong Kong, this research develops a big data-probability (BD-P) model to estimate construction waste composition based on bulk density. Using a saturated data set of 4.27 million truckloads of construction waste, the probability distribution of construction waste bulk density is derived, and then, based on the Law of Joint Probability, the BD-P model is developed. A validation experiment using 604 ground truth data entries indicates a model accuracy of 90.2%, Area Under Curve (AUC) of 0.8775, and speed of around 52 s per load in estimating the composition of each incoming construction waste load. The BD-P model also informed a linear model which can perform the estimation with an accuracy of 88.8% but consuming 0.4 s per case. The major novelty of this research is to harmonize big data analytics and traditional probability theories in improving the classic challenge of predictive analyses. In the practical sphere, it satisfactorily solves the construction waste estimation problem faced by many waste management facility operators. In the academic sphere, this research provides a vivid example that big data and theories are not adversaries, but allies.

Automated summary

By combining big data analytics and traditional probability theories, this research developed a BD-P model for estimating the composition of construction waste, achieving an accuracy of 90.2% and a speed of around 52 seconds per estimation. The study successfully addressed the challenge faced by waste management facility operators in estimating construction waste composition.