<p>Identifying geochemical anomalies associated with mineralization by analyzing geochemical survey data is essential to discovering new mineral deposits in mineral exploration. Two representative approaches for mining geochemical survey data are supervised and unsupervised recognition methods. These methods focus on geological knowledge extracted from data, including known mineralization information and potential mineralization-related anomalies, respectively, and ignore the geological knowledge obtained from mineral deposit models. Consequently, they face challenges in adequately integrating multifaceted geological knowledge to identify geochemical anomalies associated with mineralization in complex geological environments. Here, we proposed a novel model, named geological knowledge interacting reinforcement learning, for identifying geochemical anomalies linked to mineralization. This model integrated knowledge from both data and geological models into the reward feedback mechanism of reinforcement learning, enabling deep interaction between geological knowledge and artificial intelligence models. By utilizing the trial-and-error nature of reinforcement learning, multifaceted geological knowledge was synthesized to identify the mineralization pattern in complex geological environments. This novel model offers a prototype framework for comprehensively recognizing geochemical anomalies associated with mineralization and enhancing identification accuracy.</p>