A Comparison Between Several Machine Learning Methods For Multivariate Geochemical Anomaly Identification In The Helong Area, Jilin Province

Modern mineral resource prediction and quantitative evaluation is the modeling and evaluation process of complex and high-dimensional nonlinear system,which extracts and identifies the specific information in geological,geophysical,geochemical and remote sensing observation data,and predicts the spatial distribution of mineral resources based on this.The ore prospecting model based on single information source has its limitations.Only by comprehensively analyzing and synthesizing observation data from multiple sources,can mineral resources exploration and quantitative evaluation be better completed.Drainage sediment survey is an important data source for regional exploration geochemical research.The programs for multivariate geochemical anomaly detection with one-class support vector machine,isolation forest,continuous restricted Boltzmann machine and local outlier factor were developed based on the Python source codes of Sklearn.The geochemical anomalies were extracted from the stream sediment survey data of 1: 50000 scale collected from the Helong area,Jilin province which is the selected study area.By using the spatial locations of known deposit and mineral occurrences in the study area as the ‘ground truth' data,the ROC curves of the four algorithms were plotted and the AUC values were computed for comparing the performance of the four algorithms in geochemical anomaly detection.In order to better analyze the differences and scope of application of several methods based on machine learning theory to multiple geochemical anomaly identification.The result of research show that:(1)all of the four machine learning algorithms can effectively identify multiple geochemical anomalies,and the extracted multiple geochemical anomalies have significant spatial correlation with the known ore points,(2)in terms of data processing time,isolation forest algorithm is the best,followed by one-class support vector machine,and the longest is the continuous restricted Boltzmann machine,(3)among the four machine learning methods,the continuous restricted Boltzmann machine has the best recognition effect for multiple geochemical anomalies,followed by one-class support vector machine,and the local outlier factor is the worst.