BP神经网络在化探数据分类中的应用

Application of BP neural network in the classification of geo-chemical survey data

BP神经网络作为一种具有高度非线性映射能力的计算模型,具有优良的非线性逼近能力,在化探数据处理、矿产资源潜力评价等地学应用中,很多问题正是具有高度的非线性的。BP网络可自动模拟各种成矿因素之间的自然关系,进行全局优化搜索,减少人为干预,提高资源预测的准确率。BP网络具有简单易行、并行性强等特点,目前仍是多层前向网络的首选算法。以新疆东天山的化探数据对BP神经网络进行测试,分别以东天山地区的典型金矿、典型铜矿做矿床规模和类型的分类。测试结果表明,改进的BP神经网络收敛速度快,具有较高的学习效率,可以很好地对化探数据进行分类。

As an calculation model with high non-linear mapping ability, BP neural network has excellent non-linear approximation. When dealing with geo-chemical survey data and mineral resources potential assessment, many problems have non-linear features. In forecast, ANN can establish a non-linear reflection relation between the input and output, then automatically simulate the natural relation between the various mineralization factors and carry out the whole optimal searching, thus reducing the human intervention and improving the accuracy of the resource forecast. BP neural network is easy to implement with strong concurrent characteristic. At present, it is the first choice in terms of forward network. In the paper, geo-chemical survey data of East Tianshan Mountain was applied to be tested. We use the typical gold deposit and copper deposit of East Tianshan Mountain to classify the deposit scale and deposit type. The result shows that improved BP network has good convergence and high study efficiency and it can ideally classify the geochemical survey data.