Function Analysis Of Geological Big Data And Classification Algorithm Research

In digital geological science research,geological big data has become an indispensable basic material for geospatial analysis.Since the scientific essence of geological big data analysis is to extract geological variables in the sea of data that have significant contributions or important effects on geological objectives,precise analysis on the connotation and denotation of geological variables constitutes the main content of the study of geological variables themselves.This is the key to ensure the reliability and accuracy of geospatial information model.According to this requirement,this paper regards the function analysis of geological big data and geological variables as the main research content,and tries to discuss the function classification,action nature and action direction of geological variables,put forward the applicable conditions,modeling criteria and target requirement of modeling analysis of geological variables in algorithm research,and theoretically clarify the precise classification basis and spatial modeling dependence of large geological data.According to this,the quality of spatial modeling can be guaranteed.In order to analyze the full effectiveness of geological variables in geological modeling,the overall evaluation of the function of variables in the model is referred to as the function of geological variables.The so-called function is a measure of the function nature and direction of geological variables and the continuous relay of variables in the whole spatial analysis.In this paper,the functional of geological variables are divided into five functions,information conversion,degree measurement,discriminatory classification,combination association and structural optimization,and they are discussed one by one.Based on the discussion of the function of geological variables,Machine learning and deep learning are applied to the classification research of geochemical observation data,and the calculation process and precision analysis are given.so as to achieve the calculation effect of functional classification of geological variables and promote the accuracy of geological considerable data.In fact,every functional study of geological variables can be subdivided,and a lot of previous research results have been accumulated.In view of the fact that classification is the preferred subject of scientific research,this paper takes the classification and discrimination function of geological variables as the key issue to discuss,and the function analysis model and classification algorithm of geological big data as the main research content.The geochemical element measurement data in Ailaoshan area of Yunnan province are taken as the research object,the methods of support vector machine,K nearest neighbor,random forest,Gradient Boosting,naive Bayes and ensemble learning classifier were respectively used to select the relatively optimal learning type and its corresponding classification effect by means of simulation experiment comparison.The main research results are as follows:1.The connotation and denotation of the function of geological data and geological variables are discussed,and the logical relationship between geological data and geological variables is discussed.The general principles,construction conditions of extracting geological variables and assigning geological variables from large geological data sets are given.2.Five functions of geological variables are discussed in detail.Mathematical models are established for each function.The definition scope and application premise of each function model is established from the perspective of quantitative analysis.3.In the application case analysis,the classification of geochemical observation data in Ailaoshan area of Yunnan province was studied,hierarchical clustering and k-means clustering were conducted using two unsupervised learning methods.It was found that the calculated results of the sum of squares of deviations were relatively consistent with those of the k-means clustering method,that is,there were three kinds of classification results.The results indicate that it is reasonable to divide the geochemical data into three categories.4.Classification criteria based on hierarchical clustering and k-means clustering,continue to establish classification model of "high precision",respectively,with different kernel function of support vector machine,k nearest neighbor,random forests,Gradient Boosting,naive Bayes method,through the repeated simulation analysis,obtained random forest the highest classification accuracy of the conclusion,reached 99.83%;The classification effect of naiveBayes was also reached 97.74%.It shows that the random forest is preferable to other methods in geochemical data classification.In order to increase the classification accuracy of data volume,this paper proposes an integrated learning taxonomy that integrates “multiple classifiers”.In the experimental analysis,three integrated learning models were constructed,including the fusion model based on random forest,support vector machine and neural network.The classification accuracy is up to 99.83%.5.Deep neural network,Elman neural network and Jordan neural network were used for data classification.The classification results of four hidden layers were employed by deep neural network method,with the highest accuracy of 99.65%.Four hidden layers were obtained by Elman neural network method,and the classification accuracy was 99.3%.Six hidden layers were obtained by Jordan neural network method,and the classification accuracy was 99.48%.Methods research shows that the classification effect of deep neural network is better than that of Elman and Jordan neural network.