Modelling the three-dimensional (3D) geological attributes of mineral deposits is fundamental for exploration, economic evaluations, and operation of mining projects. Despite the economic and technical importance of mine-scale geological models, they are traditionally build from visual core logging observations collected by numerous geologists with different levels of experience. Consequently, 3D geological models in most mining projects exhibit large geological uncertainty, are rarely reproducible or quantitatively auditable, and significantly rely on subjective expert opinion.Machine learning (ML) specializes in extracting and improving knowledge from data by combining mathematical, statistical, and computer sciences. Geological models can be improved through the implementation of ML techniques to analyse different types of datasets generated in mining operations, including geochemical data. These techniques use algorithmic modelling to learn a task from a dataset, such as discrimination of lithology and alteration. The outcomes of ML can approach or outperform human learning, for instance, the traditional visual drill core descriptions. Geological models build through ML are reproducible, and the model performance and uncertainty are quantifiable.Multi-element geochemistry commonly represents the largest and most reliable spatially distributed dataset in mining projects. Traditional geochemical research has contributed numerous templates to discriminate lithology and alteration and improve geological modelling. However, template-based models represent generalizations of geological attributes that may not be adequate in a mining project. In addition, those templates are designed for total and near-total digestions; consequently, they are not applicable to weak digestions such as aqua regia. A ML approach expands geochemical analysis beyond generalized templates and traditional stoichiometric-based models, such as molar ratios, because a total digestion is not a pre-requisite. More importantly, the ML outcomes represent custom-made models that capture the uniqueness of the mineral deposit subject to geochemical analysis. Accordingly, unsupervised and supervised learning workflows are discussed in the context of modelling the 3D geological attributes of mineral deposits. CITATION: Ordez-Caldern, J C, 2019. Applications of machine learning to model 3D geological attributes of mineral deposits using multi-element geochemical data, in Proceedings PACRIM 2019, pp 4042 (The Australasian Institute of Mining and Metallurgy: Melbourne).