Abstract: Today, with rapid development of computer science and technology, ever more geologists are learning and applying big data based research methods, as it has become evident that many geological problems can be solved or clarified by analyzing correlativity of geological data. Here, we used multivariant statistical analysis method to mine the 1∶200000 scale geochemical survey data of stream sediments from western Qinling district, and established a series of quantitative optimization models of target areas for regional Au prospecting. These series of optimization models increased the accuracy for Au metallogenetic prediction in the study area (up to 30% Au prediction accuracy in randomly selected areas); meanwhile, they predicted the elemental composition of Au ore with high agreement with the theoretical value based on Au metallogenic geochemistry. We quantitatively evaluated each element according to its predicting power in all predication models to provide quantitative basis for further research on the genesis and controlling factors of Au ore deposit. Our results demonstrate that massive geological data possess great research potential which can only be exploited by applying big data and quantitative analytical methodologies. At the same time, it fully proved that the feasibility and necessity of quantitative optimization selection of geological research and exploration target area is realized by “identifying the relationship between data and replacing the causal relationship between things”.

Key words: geochemical prospecting, big data, data mining, series of models, regional prospecting target areas, quantitative optimization selection