Genetic Modeling of GIS-Based Cell Clusters and Its Application in Mineral Resources Prediction

This paper presents a synthetic analysis method for multi-sourced geological data from geographic information system(GIS). In the previous practices of mineral resources prediction, a usually adopted methodology has been statistical analysis of cells delimitated based on thoughts of random sampling. That might lead to insufficient utilization of local spatial information, for a cell is treated as a point without internal structure. We now take"cell clusters", i. e., spatial associations of cells, as basic units of statistics, thus the spatial configuration information of geological variables is easier to be detected and utilized, and the accuracy and reliability of prediction are improved. We build a linear multidiscriminating model for the clusters via genetic algorithm. Both the right-judgment rates and the in-class vs. between-class distance ratios are considered to form the evolutional adaptive values of the population. An application of the method in gold mineral resources prediction in east Xinjiang, China is presented.