Geovariances gives a technical presentation at MMME’2025 Paris
Discover how Geovariances' consultants are bridging geostatistics and Machine Learning to solve complex challenges in mineral resource modeling – attend their talk and learn more.
Geovariances expert consultants, Pedram Masoudi, Jean Langanay, and Roberto Rolo, are to present a technical paper at the 12th International Conference on Mining, Materials, and Metallurgical Engineering (MMME’2025) in Paris (August 2025) on :
Coherent modeling of mineral grades and zones by coupling cokriging and Support Vector Machine
Join them to explore how combining geostatistics and Machine Learning can overcome key challenges in mineral resource modeling.
This talk introduces a hybrid workflow that improves classification performance on complex categorical datasets while preserving data integrity during interpolation.
Don’t miss the case study on a synthetic porphyry copper deposit.
Abstract – The geostatistics and Machine Learning theories originate from statistics but are rooted in different applications. They are sometimes considered competing theories, sometimes complementary. This article’s latter perspective is a foundation that seeks to use them jointly to address shortcomings in mineral resource modeling. The theory of geostatistics provides methods for a robust spatial modeling of mineral resources from univariate and multivariate datasets. However, it requires considerable effort and experience to generate a coherent model when the dataset contains many categories, whether with a single categorical variable or due to the crossing of two or more categorical variables. Machine learning could address this limitation by establishing classification rules between continuous and categorical variables in the space of drillholes. The proposed workflow is applied to a synthetic porphyry copper dataset to verify and illustrate its performance in a multivariate application. The dataset comprises five mineral grades across five mineral zones. The concluding remark is that the choice of geostatistical interpolator should be made cautiously, not altering the input core data’s statistical distribution (variance and dimension support change) in the drillhole space while interpolating to the block space.