Prediction of Prospecting Target Based on Selective Transfer Network

In recent years, with the integration and development of artificial intelligence technology geology, traditional geological prospecting has begun to change intelligent prospecting. Intelligent mainly uses machine learning predict target area by mining correlation between variables metallogenic characteristics, which usually requires a large amount data for training. However, there are some problems in actual research, such as fewer sample irregular features, affect accuracy reliability prediction. Taking Pangxidong study Guangdong Province an example, this paper proposes deep framework (SKT) prediction based on selective knowledge transfer carries out research geochemical Pangxidong. The features different scales captured dilation convolution, weight parameters source network selectively transferred networks training, so increase generalization performance model. A number experimental results show that method obvious advantages over other state-of-the-art methods areas, effect samples mines is greatly improved, can effectively alleviate small areas