Geochemical Pattern Recognition: Three New Geochemical Pattern Recognition Methods And Their Application

Pattern recognition plays an important part in exploration geochemistry data interpretation. In a word, pattern recognition is classification. What is called geochemical pattern recognition proposed in this paper is a method using statistic pattern recognition way to solve exploration geochemistry classification problems.As we know, the main idea of geochemical data processing and interpretation is classification. For example, in exploration geochemistry various univariate distinguishing abnormity from background methods in common use is actually against to classification problems. A majority of multivariate statistical methods in common use, including discrimination analysis, cluster analysis, factor analysis, R-Q mode factor analysis and correspondence analysis et al is also against to classification problems. As the final target of exploration geochemistry which needs to recognize mine or not, and mineralization type is also classification problems.Traditional pattern recognition methods have successfully applied in exploration geochemistry. However there are many problems in these methods. The first problem is in R-mode cluster analysis the minus correlation among variables cannot be manifested. Recognizing this relation efficiently is very important to result interpretation, otherwise, it will lead to mistake and unilateral interpretation. The second problem consist in the univariate methods, many researchers have pointed out the irrationality of the method even consider that these methods is misleading. The third problem is the research target of the methods in common use is singularity, these methods research variable types (R-mode analysis) or sample types (Q-mode analysis) respectively, and do not systematically study the duality of the two, however the duality correlation is very important to geochemical data interpretation. The fourth problem is the classification of large data set is still not well solved. For example, in the commonly used C-means cluster analysis, the number of the classes needs to be subjectively chosen, and it is not so convenient in applications, so the method is hard to be generalized in applications. It is worth to point out that the problems above still do not bring people's fully attention, and the concept of"geochemical pattern recognition"is firstly proposed in this paper.Against to these problems, the target of this paper is: First, try my best to describe the concept of geochemical pattern recognition, typical hierarchical and non- hierarchical classification and their application systematically. Second, propose anti-R-mode cluster analysis. Third, introduce and propose a few other classifications, including hierarchical correspondence cluster analysis, semi-hierarchical correspondence cluster analysis and R-Q mode C-means cluster analysis. These methods can simultaneously solve classification of large data set and the duality correlation between R-mode and Q-mode analysis. Fourth, take the practical data as an example to introduce geochemical pattern recognition method using in mine area and regional geochemical exploration, and discuss their application effect.As the concept of geochemical pattern recognition is first proposed, this paper systematically introduces its significance and basic concept. In order to systematic discussion and point out the origin of the new method, this paper introduces some classical methods briefly, including hierarchical cluster analysis and C-means cluster analysis.The point of this paper is to narrate some new statistic pattern recognition methods. The anti-R-mode cluster analysis the author proposed is a new method to solve the minus correlation hard to express among variables in R-mode cluster analysis. Semi-hierarchical correspondence cluster analysis, extracts the main advantages of correspondence analysis, hierarchical and non-hierarchical cluster analysis, and unifies the R- and Q-mode cluster analysis of large data set. R-Q mode C-means cluster analysis firstly develop in this paper, unifies the main advantages of R-Q mode factor analysis and C-means cluster analysis, in order to solve the classification of large data set and the duality correlation between R-mode and Q-mode analysis as well. This paper introduces the basic thought, mathematical technique, application approach, calculation instance of these methods in detail.As another point, this paper introduces a new pattern recognition method using in mine area and regional geochemical exploration, and discuss their application effect. In Shancheng Gold Mine we use anti-R mode cluster analysis to recognize the relationship among the altitude of the constructional surface and much geological, geochemical information, consequently know geological structure control gold deposit correctly, and provide important information for structure geochemistry and further exploration.In Jishan Gold Mine, we use semi-hierarchical correspondence analysis to recognize the geochemical types of the samples, geochemical zoning, metallogenic pattern of industrial ore body and exploration target. In Jiudian Gold Mine we use semi-hierarchical correspondence analysis to reveal various geochemical characters, and successfully forecast the hidden ore body from the low content area.In the shallow covered area around Tahe town in Heilongjiang province, we use semi-hierarchical correspondence analysis to interpret the regional geochemical data, and recognize the regional lithology, structure, types of the deposit and exploration target. We also attempt to use R-Q mode C-means cluster analysis in this area, and gain certain effect. As time was limited, the test needs further research.Various theory research and application effect show the important significance of these methods. In theory, these methods solve some problems in multivariate statistical analysis, or provide certain approach to these problems, and make contributions to multivariate statistical analysis and mathematical geology. In application, these methods can provide simple steps and only one figure to show main information of the multivariate data set, to reveal geology and exploration problems which cannot discover with ordinary methods, and make contributions to exploration geochemistry.