Proposal for a Machine Learning Model to Improve Hauling Productivity in an Open-Pit Mine

Abstract

This proposal outlines a machine learning-based approach aimed at improving productivity in haulage operations within open-pit mining. Since hauling accounts for up to 60% of total operational costs, predictive models that enable early intervention and optimization are of strategic importance. The proposed methodology involves the use of Gaussian Mixture Models (GMM) for data preprocessing and Random Forest algorithms for predictive modeling, complemented by ensemble techniques such as Gradient Boosting and XGBoost. The model is expected to be trained and evaluated using historical and real-time operational data, including variables such as loading time, truck availability, material type, and travel distance. Evaluation metrics such as MAE, RMSE, and 𝑅 2 will be used to assess predictive performance. The aim is to build a framework that enables early warnings of productivity deviations and supports real-time decision-making. This research seeks to contribute to Mining 4.0 through the development of an interpretable and scalable tool for haulage optimization in real operational settings.