Metallogenic Prediction Based On Fractal Theory And Machine Learning In Duobaoshan Area,Heilongjiang Province

The Heilongjiang Duobaoshan area is one of the important mineralization areas in China,with excellent geological conditions for mineralisation,containing a variety of rich mineral resources such as copper,molybdenum,gold,silver and iron,etc.Large and medium-sized deposits have been discovered with numerous mineralized occurrences.However,the complexity of the geological conditions and the multi-phase mineralization have increased the difficulty and uncertainty of exploration in the area,posing challenges to further exploration and search.This paper focuses on some works including geological structure interpretation,remote sensing alteration information extraction and chemical anomaly information extraction in the Duobaoshan area of Heilongjiang,and uses fractal theory to carry out a comprehensive analysis.The multisource information database for mineral prediction was established,and finally the prediction model for mineralization based on machine learning algorithms was constructed to delineate the prospective areas for mineralization.The conclusions are as follows:(1)Fracture structure interpretation and fractal analysisBased on Landsat-8 OLI remote sensing images,geotectonic interpretation was carried out using interpretation markers,and image fusion processing with optimal band selection was employed to interpret geological structures.Fractal dimension values of regional structures were calculated to analyze fractal characteristics.The results showed that most of the ore points were located in the range of structural fractal dimension values ranged from 1.4 to 1.8,and the fractal results were consistent with the spatial distribution of metallogenic structures.High-value areas were found in the eastern part of Yongxin gold deposit,Nanshan gold deposit of Huolongmengou,and Yezhugou molybdenum deposit.(2)Remote sensing erosion anomaly extraction and fractal analysisBy using "Interference removal + band ratio + principal component analysis + fractal thresholding segmentation" to extract and analyze alteration anomaly information,the remote sensing alteration information was found to exhibit spatial fractal characteristics.Fractal values of iron stain alteration ranged from 0.9 to 1.7 and showed a certain spatial correlation with iron stain alteration.Fractal high-value areas were consistent with dense areas of iron stain alteration information.Most of the hydroxyl alteration dense areas were also located in their fractal highvalue areas,which were mainly found in the Yongxin gold deposit and the Huolongmen area.(3)Geochemical anomaly information extraction and fractal analysisExploratory data analysis was conducted on the main mineralization-indicating elements of the soil geochemical data.Isometric log-ratio transformation was used,and robust principal component analysis was applied to explore the correlation and co-association patterns between elements.The results of double-labeled principal component map interpretation were combined to clarify the significance of elemental combinations on mineralization.The energy S-A fractal model was used to decompose the composite anomaly for robust principal component analysis combination,highlighting the true spatial distribution of mineralogic anomaly information in the Duobaoshan area.(4)The mineral resource forecasting based on machine learning methodsThrough sample equalization,and the prediction layer of the multi-source mineralization model was created by combining the results of structures and alteration fractal information and chemical anomaly information.Hyperparameter search was performed using grid search.Based on the prediction favorability of both models,the geological mineralization background,structures and remote alteration fractal information,and geochemical anomalies were integrated using the CA fractal model to divide the mineralization potential.Eight three-level metallogenic prospective areas were identified by integrating the geological background,structures and alteration fractal results,and geochemical anomalies,including two A-class,four B-class,and two C-class metallogenic prospective areas.