Multi-Scale Progressive Mineral Prospecting Prediction of Hidden Manganese Ore Deposits Based on Geological Big Data in Northeastern Guizhou, China

Exploration for hidden ore bodies is important and requires new prospecting methods. However, predicting the presence of deep hidden mineral resources is difficult because of the large prospective area and the numerous factors influencing underground geological structures and mineral resources; in addition, available information on geological structures and their relationships and changes over time is incomplete. This study focused on a region with typical Mn ore deposits in the Upper Yangtze Block, northeastern Guizhou Province, South China. A geological model based on basic geology and ore mineralogy knowledge was combined with model-free prediction based on the fourth paradigm of scientific research—which uses the multi-scale progressive technical strategy from 1P to 5P (xP refers to various scales of mineralization potential prediction)—to generate a data-and-model-driven big data mineral prospectivity model (BDMPM) that organizes big datasets and big data chains to predict Mn ore mineralization potential. The validation of BOA-AdaBoost in the Songtao Mn ore-concentrated areas demonstrated that the prediction model at the 3P scale achieved an accuracy of 0.936, precision of 0.948, recall of 0.925, F1 score of 0.928, Kappa of 0.862, and AUC of 0.967. This also indicated that the proposed BDMPM can provide technical support for the prospecting and exploration of hidden Mn ore deposits.