Combining neighborhood component analysis with dictionary learning algorithms to improve the performance of the dictionary learning models for geochemical anomaly detection

Abstract

In geochemical exploration, a geochemical anomaly detection model is usually established to describe the population distribution of geochemical data, and samples that do not conform to the model are identified as geochemical anomalies. Because the establishment of a geochemical anomaly detection model does not make use of the relationship between geochemical elements and mineralization, the performance of geochemical anomaly detection model for mineral exploration targeting is affected to a certain extent. For this reason, neighborhood component analysis and dictionary learning algorithms were combined to detect geochemical anomalies associated with gold mineralization in the Chengde area in Hebei Province, China. Neighborhood component analysis was used to transform geochemical data from the input space into the neighborhood component space to enhance the separability between the geochemical anomalies associated with gold mineralization and the background. Dictionary learning models for geochemical anomaly detection were established in the neighborhood component space. The performance of the dictionary learning models established in the neighborhood component space was compared with that of the corresponding models established in the input space in geochemical anomaly detection. The results show that the dictionary learning models established in the neighborhood component space are superior to the corresponding models established in the input space in geochemical anomaly detection. In addition, there is a strong consistency between the mineral exploration targeting results and metallogenic characteristics of the study area. Therefore, combining neighborhood component analysis and dictionary learning algorithms can improve the performance of the dictionary learning models in geochemical anomaly detection.