Extraction Of Geochemical Multi-Element Associations Based On FastICA And Its Application

Identifying anomaly element and geochemical element associations is the important fundament of figuring out foundation geology, metallogenic problems and mineral resource evaluation by means of geochemistry. This problem used to be solved by direct comparative method and statistical analysis method, the former determines the cause of the high content of element associations by comparing the metallogenic belt geochemical element associations high content area to the distribution of mineralized bodies;the later studies the geological significance of element associations in statistical methods. Though both them are of indicator function in geochemistry, they cannot meet the demand of searching for deep hide and difficult identified mineral deposit, following the improvement of the degree in mineral exploration and the shrinking of the outcropping and recognizable ore deposits. Consequently, we have to explore and apply new theories and new methods in terms of mineral resources prospection and geological prospecting work.ICA is a newly developing method in signal analysis and processing. It takes higher-order statistics belonged to non-normal distribution as research object; its aim is to get linear transformation of data which is statistically independent and non-Gaussian distribution. The characteristic of the distribution of geochemical elements just is highly non-linear. So, ICA satisfies the essential prerequisite of geochemical data processing.The paper had rebuilt original model in order to meet the characteristic of geochemical element associations.To begin with, geochemical survey data of single element are transformed into "mixed signals" of new model. After data preprocessing, objective function should be determined with the help of negentropy maximization and solved optimal value by newton iteration algorithm. Finally, we must do an inverse transformation so as to access element associations with "maximum energy". The result derived from the comparison between new method and traditional method, revealed that geochemical element associations made by FastICA can truly reflect study area situations. It seems that ICA has a big advantage over the traditional one in digging high correlation information.