Abstract

Geological mapping, as the fundamental core of geological research and investigation, provides indispensable basic understanding and exploration data for mineral prospecting, disaster prevention and control, and other related fields. The emergence of big data algorithms and models, such as machine learning and deep learning, has brought new assistance to the task of geological mapping. By utilizing these advanced technologies, long-term accumulated exploratory data has been deeply mined and analyzed, significantly enriching the information obtained through conventional mapping methods and laying a solid foundation for subsequent research. However, challenges faced by machine learning in the field of geological mapping are gradually being recognized: the vast demand for data volume, interference from irrelevant information, the “black box” problem, limitations in computing power, algorithm applicability, and data security, among others. To address these challenges, this study takes the Duolong mineral district, Tibet, China, a region with a high degree of exploration, as an example, assuming it to be a blank area without mapping, and attempts to integrate the light gradient boosting machine (LightGBM) algorithm into the field geological mapping process. Firstly, based on remote sensing data, the distribution of alteration was identified to design the initial mapping route. Subsequently, geological units along the route were labeled, and the model was trained using geochemical sampling point data and remote sensing data as feature inputs. During the model prediction stage, the probability distribution obtained through the Softmax function was utilized to guide the subsequent design and planning of field mapping routes. After five iterations, based on field mapping that covers 20% of the entire area, 90% of the lithological units were successfully predicted. This study explores an effective combination of machine learning algorithms with field geological mapping that establishes a new method for field geological mapping based on machine learning. It not only improves the efficiency and accuracy of mapping but also provides a new strategy for balancing geological work with structured data acquisition.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call