The Methods Of Constructing And Prospecting Based On Mineral Deposit Knowledge Graph

With the advent of the information age,geological big data is faced with problems such as huge data volume and low mining efficiency.Knowledge-driven big data analysis methods and theoretical research in the geological field are the strategic focus of contemporary geological knowledge research.It can effectively organize and represent geological big data.This paper proposes a method of constructing a geological knowledge graph with ore deposits as the ontology.As a comprehensive product of geological processes,ore deposits cover most of the geological researches in terms of their metallogenic materials and their sources,metallogenic environment and metallogenic processes.Content,building a knowledge graph with mineral deposits as a unit can comprehensively and systematically organize and develop geological data.The main research contents and results of this paper are as follows.(1)In this paper,the ore deposit knowledge graph is constructed from the two parts of "ore deposit prediction model-typical ore deposit";And according to the idea of "top-down,bottom-up,subdivisions meet,and stratification",the knowledge of ore deposits is represented,and the integrity and difference of the metallogenic characteristics of the ore deposits are shown from a deeper level.(2)The ore deposit entity extraction model based on Bert-Bi LSTM-CRF and the Bert-based relation extraction model were trained by using the labeled ore deposit data set to realize the intelligent extraction of ore deposit knowledge graph triples;and based on the constructed graph,ore deposit knowledge graph management system was developed to realize the visual display of ore deposit knowledge,the analysis of geological elements based on graph algorithm,spatial index and other functions,and to provide convenient information services for retrieving ore deposit knowledge.(3)This paper takes the Zhaishang-Mawu gold deposit area in Gansu as the research area,and constructs the basic geological knowledge graph of the research area,as well as the Mawu Gold Mine,Zhaishang Gold Mine,Liba Gold Mine and Suolong Gold Mine in the mining area.Luerba Gold Mine,Xinzhuangli Gold Mine,Guojiagou Gold Mine,Hualingou Gold Mine,Zhuzigou Gold Mine,and nine typical gold mines.The division of prospecting areas,the analysis of genetic types and the comparison of metallogenic characteristics provide new perspectives and clues for prospecting and prediction.(4)Based on the triple representation of "entity-relationship-entity" in the knowledge graph,the centrality algorithm,community detection algorithm and similarity algorithm are combined to summarize and analyze the prospecting clues in the study area in detail,research and realize the knowledge graph of mineral deposits.The prospecting prediction method combined with the image classification technology gives semantic information to the geological graph in the form of a vectorized triad of knowledge graph,and embeds it in the convolutional neural network for prospecting prediction,and has achieved good experimental results.