为了进一步确定量化矿床模式是否令人满意,本文对利用矿物成分概率神经网络法进行矿床分类做了检验。典型类型矿床的矿石和蚀变矿物成分存在与否被用来训练神经网络。为了降低要考虑的矿物种类,只利用至少一类矿床中矿物区占20%以上的那些矿物。这一限制的好处在于单个或很少出现的矿物不会影响到正确分类结果。 In order to further determine whether the quantitative deposit pattern is satisfactory, this paper tests the mineral deposit classification using the probabilistic neural network method of mineral composition. The presence or absence of ore and alteration minerals in a typical type of deposit is used to train neural networks. In order to reduce the types of minerals to be considered, only those minerals which account for more than 20% of the mineral area in at least one type of deposit are utilized. The benefit of this limitation is that a single or infrequent mineral does not affect the correct classification result.