Randomized trial demonstrates accurate construction waste prediction in various data sources, indicating improved resource management.
Key Points
This research aims to develop a robust model for predicting construction waste by integrating various machine learning methods with grey relational analysis.
Developed a grey-relational ensemble model combining backpropagation networks, convolutional neural networks, and long short-term memory networks with random forest as the meta-learner.
Utilized datasets from Eurostat and the U.S. Environmental Protection Agency.
Evaluated model performance using metrics like mean absolute error and R2.
Achieved a mean absolute error of 5.4 tonnes and a mean squared error below 12.6 tonnes².
Obtained an R2 of 0.913 and a waste processing response time within 0.8 seconds.