The activity from which construction waste arises includes new construction, renovation, and demolition. Renovation waste was traditionally considered as trivial but in recent years, it has gradually come into the focus of construction waste management (CWM). Waste auditing prior to renovation (termed ‘pre-renovation auditing’ or PRA) provides a departure point for good CWM. The core of PRA is accurately predicting renovation waste generation. Benefiting from a valuable dataset containing 351 building renovation projects in Hong Kong, this research aims to develop a robust renovation waste estimation approach. By using machine learning regression, a model containing several easy-to-access features was developed. By simply inputting the feature data (including renovation work type and cost; building type, year, height, and floor area; as well as floor height and the number of renovated floors) into the model, the renovation waste generated from the project can be reliably estimated. Validation experiments indicate that the method has a root mean squared error (RMSE) of 141.52, mean absolute error (MAE) of 79.69, and R-square of 0.83. Comparative experiments show that our method performs better than prevailing ones as reported in the literature. This study thus contributes a novel waste prediction model for pre-renovation waste auditing. With proper modifications, the model can be applied to other regions with different building and renovation features. Future research is recommended to further explore advanced waste estimation methods by integrating new technologies for finer CWM.