Application Study On Qualitative Information Processing Technology

The statistical analysis on information is frequently encountered in solving the actual problems, and certain qualitative variables often play decisive role in the whole statistical analysis process, such as qualitative factors like lithology and structure etc. that must be taken into account in mineral resources evaluation. The study & application of processing and analysis methods of qualitative information has presented more and more important theoretical importance and practical value.Based on two actual cases in mineral resources evaluation and educational statistics evaluation, this paper combines the artificial intelligence ideas such as modern fuzzy mathematics theory and artificial neural network into traditional statistical analysis method of qualitative data, and carries out in-depth application study on processing technology of qualitative information. The paper comprises two parts, the first part presents two kinds of evaluation and prediction methods based on characteristic analysis and resilient back-propagation neural network designed for the qualitative data in actual mineral resources evaluation; Combination of linear algorithm characteristic analysis with non-linear algorithm characteristic analysis is used in the mineral resources evaluation, providing a new route for solving mineral resources evaluation problems. The second part presents three kinds of analysis & evaluation. methods of multi-dimensional qualitative data based on dual scaling, fuzzy classification histogram and multi-layer feed-forward fuzzy neural network, designed for multi-dimensionalqualitative data in educational statistics evaluation; These methods take into full consideration of the fuzziness of certain qualitative variables' meaning and transition among obtained statistical data, adding new methods of modern theory to processing technology of multi-dimensional qualitative data.The study on theory and methods of regional mineral resources prediction is one main research contents of mineral resources evaluation, engineered to solve the prediction on resource potential and possible occurrence spatial position of resources in a certain geologic range of certain mineral resources. According to the synthetic information of geology, geophysical and geochemical prospecting in Guizhou Province provided by Guizhou Geologic Survey Institute, we finalize the variable weight by the help of the characteristic analysis model deducted by product matrix vector length method (quadretic sum method) through several model test screenings. The first part of paper is the positioning prediction & evaluation on the region's gold mine. It starts with division of modules and withdrawal of geologic variables, takes then the known Au concentrated regions as the model module and Au abnormally concentrated regions as prediction module, uses characteristic analysis positioning prediction model to find deposits of same variety and characteristics by the help of degree of association of all modules, and based on the modular degree of association, carries out positioning prediction evaluation on Au abnormally concentrated regions in Guizhou Province through mineralization probability estimation of different modules and ultimately, selects 52 target areas by Grade I and II positioning prediction as primary working module with perspective, among which. The evaluation offers new basic information for the further quantitative prospecting & prediction of gold mine in Guizhou Province.Thanks to the eminent features of artificial neural network, it is widely used in various fields. Taking into consideration of artificial neural network's unique self-learning capability and non-linear information processing capability, we choose the back-propagation algorithm as the quantitative prediction tool of mineral resources as it is suitable for processing the geologic information and use matrix analysis method to give clear-cut mathematical derivation & formulation of forward propagation process of information and back propagation of errors. To overcome the limitation of basic BP algorithm in convergence rate and network accuracy, an improved BP algorithm-calculation procedure of resilient BP algorithm is developed. It adopts the network training information represented by gradient's symbol to adjust the network and set a weight renewal value; the network weight's corrected value is calculated by weight renewal value, whichresults in the great improvement of network convergence rate and calculation accuracy. This paper takes the Au concentrated regions with proven reserves in Guizhou Province as the prediction model module, carries out classification prediction resilient by BP algorithm on 52 Au abnormally concentrated regions module(prediction module) rated as Grade I and II through positioning prediction in Guizhou Province, obtaining the working target areas of two groups of large-size gold mines in different geologic positions of Guizhou Province and realizing the mineral resources prediction evaluation combining the characteristic analysis in linear theory and neural network in non-linear theory. It has certain theoretical instruction meaning and actual economic value.The second part addresses the educational statistics evaluation. Recognizing the fuzziness of qualitative variables' meaning and transition among statistical data, it combines the modern fuzzy theory, artificial neural network with traditional dual scaling statistical analysis method, carries out study on classification characteristics of multidimensional qualitative data of various classification variables from different angles; three different kinds of processing technologies are adopted:1. Multidimensional dual scaling method based on dual scaling. To explore the impact on general structural features of samples imposed by classification using various classification variables, the author sorts firstly the complex qualitative information into multidimensional datasheet, classifies appropriately the samples according to different classification variables through design matrix, determines the variable weight, sample score and factor axis corresponding to bigger characteristic value through dual scaling as well as the actual meaning of factor axis, draws finally scattergram of different sample subset on factor plan determined by factor axis, analyzes & summarizes the statistical rule and conclusion of among different variables, samples, samples and variables, and among the substructures of classified samples.2. Fuzzy classification histogram statistical analysis method based on fuzzy classification theory. In the process of study on multidimensional dual scaling method, it is noticed that the actual meaning of factor axis determined by certain factors has fuzziness, it is therefore necessary to introduce fuzzy classification theory as we have to incorporate it in the statistical analysis of multidimensional qualitative data. Based on the multidimensional dual scaling statistical analysis, to analyze more accurately the sample subset's...