Research on the method of expanding the mineralization data based on generative adversarial networks

In the domain of mineral resource exploration, a pronounced imbalance often exists between the number of mineralized samples and the number of samples that are found to be non-mineralized. To address this imbalance, this paper proposes an AC-CTGAN method for expanding mineralized samples based on generative adversarial networks, utilising the northeast Guizhou manganese mining area as a case study. The findings of the research demonstrate that the data generated by AC-CTGAN exhibits excellent performance in terms of spatial distribution and bears a strong resemblance to the feature distribution of real data. The average cross-validation accuracy of the prediction model trained using the expanded dataset is improved by 8\%, and the prediction ability of the model is significantly enhanced.