Prediction Of Regional Mineral Resource Potential Based On Mineralization Case-Based Reasoning Model

The demand of mineral resources was increasing sharply due to the development of modern industry.The deposits that are deeper,less recognizable,less developed and utilized have become the new exploration and development goals as a result of the mines which are shallow and easier to mining are nearly exhausted.In addition,indirect mine-prospecting methods such as digital mines and smart mines have gradually replaced direct mine-prospecting methods such as man-made mines and mechanical mines.The direct mine-prospecting methods are difficult to adapt to the current storage environment of mineral resources.Though the prospecting efficiency of indirect methods has been improved significantly,there are some problems,including low flexibility,weak scalability,and poor reuse of new results.Case-based reasoning is a branch of artificial intelligence that has been widely used in various fields.In case-based reasoning,we just need to extract the case features,and compare the new case without result to the history cases with results,then we can get a similar case which is most similar with the new case.Finally the new case can get the result from the similar one.The case-based reasoning is a very effective tool for the problems which are fuzzy,nonlinear,and hard to extract rules since it is difficult to build accurate domain models.This thesis combined case-based reasoning with geographic information technology in the problem of potential prediction of regional mineral resources,used multi-source geospatial data,fully explored the implied spatial proximity between the metallogenic units and the ore-controlling factors,built an integrated mineralization case expression model combined attributes with spatial proximity features,which is different from traditional case expression models.Then,a mineralization case similarity reasoning model was built.In addition,a prototype system of metallogenic case-based reasoning was developed,and the metallogenic case-based reasoning experiments were conducted to predict gold mineral resource potential in the eastern Kunlun area of Qinghai province.Experimental results show that,the proposed metallogenic case-based reasoning model has better gold mineral prediction accuracy over the traditional case-based reasoning.In the first test dataset,the traditional method's prediction accuracy in the high storage potential area and high-medium storage potential area is 85.81% and 89.19%,respectively.However,the prediction accuracy of the proposed method is 85.81% and 89.86%,respectively.In the second test dataset,the traditional method's prediction accuracy in the high storage potential area and high-medium storage potential area is 88.24 % and 98.53%,the prediction accuracy of the proposed method is 91.18% and 98.53%,respectively.Furthermore,the proposed method has better prediction accuracy over the weights-of-evidence model which has the accuracy of 74.40% in the gold's high-medium storage potential area.The thesis draws the following conclusions.(1)Compared with the traditional case-based reasoning model which uses only attribute features,the metallogenic case-based reasoning proposed by this thesis with the simultaneous integration of attribute and spatial proximity features has better prediction accuracy in gold mineral potential.The thesis shows that fully excavating the implied spatial proximity between the metallogenic units and the ore-controlling factors,and building an integrated mineralization case expression model with attribute and spatial proximity features to conduct metallogenic case-based reasoning is more effective.(2)Compared with the weights-of-evidence model,the metallogenic case-based reasoning model proposed in this thesis gives a more significant prediction result to gold mineral resource potential.Experimental results show that the metallogenic case-based reasoning model is more feasible for the potential prediction of gold mineral resources.