In real-world data sets, missing values are unavoidable for various reasons. These missing values are typically represented by NaNs, default placeholders, or simply left as blank entries. Depending on the extent of missing data, this can significantly reduce the performance of statistical methods. Additionally, data sets with missing values are incompatible with many machine learning techniques, including random forests, regression models, and neural networks, which rely on the assumption that all features contain complete and relevant information related to the task at hand. Geological data sets, which capture the 3D representation of a deposit using geological field observations, survey data, drill hole information, and assay grades, are no exception. In geological modelling, it is extremely rare to encounter complete data sets without any missing values. A common but simplistic approach is to exclude observations with missing values altogether. However, when a large portion of the data set contains missing entries, removing those records leads to substantial information loss. This highlights the importance of integrating effective missing data imputation techniques into the data preprocessing workflow-a process that presents several challenges (Rahm and Do, 2000). Effective imputation methods must account for naturally occurring geological patterns, such as the formation and spatial continuity of rock types, the proportions of different lithologies, and the uncertainty or potential misinterpretation introduced by incomplete data. A method has been developed that captures and reformulates a high level of correlation with the existing geological data. Performance gains using newly imputed data as input to machine learning processes are evaluated using several metrics to provide geological plausibility for the method.