摘要
Geochemical survey data is non-linear and randomness.Although the traditional method of identifying geochemical element associations anomaly takes the randomness into account,it ignores the non-linear of original data.Independent Component Analysis,however,not only requires second-order statistics of each component to satisfy Non-Gaussian distribution,but need to high-order statistics are non-correlation.It seems that this feature gives ICA a big advantage over the traditional one for digging high correlation information. This paper will launch the analysis and evaluation of multi-element association anomaly by an area in Tibet 1∶10 000 metrical soil data.