Research On Geochemical Anomaly Identification Method Based On Isolation Forest Model

Geochemical data is one of the important materials used in the exploration of mineral resources.Geochemical anomaly identification is the core content of geochemical exploration.In recent years,scholars have proposed various geochemical methods to obtain the lower limit of abnormality.The traditional method of obtaining the lower limit of abnormality is mainly based on the traditional statistical principle Assuming that the geochemical data obeys the normal distribution or the lognormal distribution,In this paper,an isolation forest algorithm is selected to construct a geochemical anomaly recognition model,which does not require the original data to meet certain conditions and directly isolates anomaliesThe isolation forest model constructed in this paper unfolds the anomaly recognition of the study area from two aspects:single element and multi-element.In single element anomaly identification,considering the possible invalid or high values of the original data to affect the model's effect,the data is preprocessed to select mineralization indicator elements,and the optimal threshold is used to identify geochemical anomalies.Taking 1:200 000 geochemical exploration data of Fusong area in Jilin Province,NE China as the example,Fe2O3?Pb?Zn?Au and Ag were selected as the indicator elements to identify geochemical anomalies,and the results were compared with traditional statistical methods.In multivariate anomaly identification,seven combinations of Au,Ag,Cu,Pb,Zn,Hg,and Sb mineralization indicator elements were selected using the Youden index method to complete anomaly identification,and ROC curve analysis technology was used to predict and evaluate the resultsSingle element anomaly recognition,it can be found from the distribution range and intensity of the anomalous area delineated by the two methods that the anomalous area delineated based on the isolation forest model basically covers the anomalous area delineated based on traditional statistical methods,and enhances the mineralization information corresponding to the indicated elements.The experimental results can visually see that the anomaly recognition results under the isolation forest model are more closely related to the known mineralization points,and the recognition effect is better,indicating that the best anomaly probability map determined by them is known to the corresponding metal deposit point.There is a strong spatial correlation between them,and the identified locations of high anomalous probability have certain predictive power for unknown mineralization information.Therefore,other areas with anomalous intervals identified based on the isolation forest model have other areas with high probability of anomaly,but no deposits have been found at present,which can further provide direction for prospecting work.Secondly,due to the characteristics of the isolation forest algorithm,in addition to the normal identification of high value anomalies,the isolation forest model can identify low and weak anomalies other than high value anomalies Traditional statistical methods often identify only high values due to the determination of the abnormal lower limit.abnormal.The results of multivariate anomaly recognition indicate that multiple metallogenic indicator elements were selected through the AUC value of metallogenic association index.The multi-geochemical anomaly recognition area completed by the isolated forest model and the known metal mineralization points in the study area have better connection relation.The ROC curve analysis technology is used to further prove the effectiveness of its mineralization prediction effect.