Challenges of Geological Prospecting Big Data Mining and Integration Using Deep Learning Algorithms

Mining and integrating geological prospecting information using deep learning algorithms (DL) has become a frontier field of mathematical geoscience. DL, which is a machine learning algorithm with multiple hidden layers, starts to be used in mining the geological prospecting big data in recent years, and there are a series of issues to be solved in this field. In this study, we took the convolutional neural network (CNN) as an example to discuss two challenges of DL on mining geological prospecting big data, which include insufficient training samples and how to construct deep learning network structure. In this study, the data augmentation methods were applied to generate training dataset, duplicating and adding noise, and a number of number of experiments were carried out for determining the optimal hyper-parameters of a CNN model for mining and integrating geological prospecting big data. A case study from Southwest Fujian Province, China, was carried out to mine and integrate the geological, geophysical and geochemical multi-source prospecting information. The results obtained by CNN can provide clues for mineral exploration in this area. © 2021, Editorial Department of Earth Science. All right reserved.