Geological Knowledge‐Guided Dual‐Branch Deep Learning Model for Identification of Geochemical Anomalies Related to Mineralization

Abstract Geochemical survey data are a type of spatial big data that play an increasingly significant role in mineral exploration. One challenge in the era of big data is how to mine geochemical data in support of mineral exploration. In this study, based on a generative adversarial network framework, we proposed an unsupervised spatial–spectrum dual‐branch deep learning method for geochemical anomaly identification, namely dual‐DL, which consists of a spatial branch and a spectrum branch. The spatial branch was constructed using the convolutional neural network and convolutional autoencoder, which can effectively capture spatial geochemical patterns and extract spatial relationships between neighboring pixels. The spectrum branch consists of a recurrent neural network that can study geochemical elemental assemblies within a single pixel. The geological knowledge was added into the model, including selecting the input order of geochemical elements and constructing the loss function of the model. A case study was conducted to recognize geochemical anomalies associated with gold polymetallic mineralization in Hubei Province, China. The results demonstrated that (a) the unsupervised dual‐DL model has superior performance in identifying mineralization related to geochemical anomalies, (b) the geological knowledge‐guided unsupervised dual‐DL model can improve the accuracy and interpretability of geochemical anomaly identification.