Study On Mineralization Forecast Of Xinjiang Hongyuntanchilongfeng Iron Mine Belt Based On Improved AdaBoost Algorithm

In recent years,geospatial information processing,machine learning and other technology have promoted the rapid development of spatial quantitative prediction and evaluation methods in the field of geosciences.The metallogenic system is a multielement and heterogeneous complex system.The cause,development and result of metallogenic events have complex features on the space.Therefore,it is difficult to accurately classify a given area as "mineral" or "non-mineral",but to give a "probability of mineral",which usually represents a fuzzy logic.The Real Adaboost classifier is an expansion and improvement of the traditional Discrete Adaboost classifier.Compared with the binary output of traditional Discre Adaboost,each weak classifier of Real Adaboost outputs a "probability of belonging to a certain class",and by mapping the probability value of 0-1 to the real number field.The meaning of "Real" can express more complicated classification logic.The main research contents of this article are as follows:(1)This paper introduces the proposal and development of Boosting algorithm,and analyzes and explains the error of traditional and Real AdaBoost.(2)By analyzing the traditional Adaboost algorithm,there are also some problems in the calculation formula of classifier weights,such as: When the classification error rate is close to 0,the weight of the classifier is positive infinity,therefore,the excessive weight affects the accuracy of the final strong classifier;Due to the difficulty or abnormal samples,when updating the sample weights,the weight of the sample that has been divided many times is too large,which affects the choice of the classifier and reduces the classification accuracy.This paper proposes an improved method based on the weight formula of the traditional AdaBoost algorithm classifier.In the improved calculation formula,it can better solve the problems in the traditional algorithm and improve the prediction accuracy.(3)A demonstration study on the application of the Real AdaBoost algorithm for the mineralization prediction of iron ore resources was carried out in the East Tianshan metallogenic belt,and good prediction results were obtained,which expanded the application field of the method and provided a working idea for metallogenic prediction in the research area.In this paper,Boosting algorithm,especially Adaboost and Real Adaboost algorithms that have a certain research basis in other scientific fields,which are applied to the forming mineralization prediction of iron ore in the East Tianshan metallogenic belt in Xinjiang,China,so as to verify the application of Adaboost related methods in metallogenic prediction The effectiveness of the study and provide a basis for research in related fields.In addition,the case study further promotes the application of the AdaBoost algorithm and its variant algorithms in multivariate information integration and mineralization prediction.