Abstract Yanshanian magmatism develops well and has obvious metallogenic specificity in the Qinzhou–Hangzhou bays of South China. Zircon, in granite, has great potential in the study of metallogenic specificity. Moreover, with the development of in-situ zircon analysis technology, a significant amount of data on zircon composition has become available in recent years. Hence, to further elucidate the distinctive diagnostic signatures of zircon signatures regarding metallogenic fertility and the preference for metallogenic types of granites, this paper exhaustively explores all possible bivariate and ternary discrimination diagrams using a set of 13 rare-earth elements with published zircon data from Yanshanian, South China. This results in 4095 binary diagrams and 12,485 ternary diagrams, ranked according to their success in classifying the training data. One-hundred binary graphs and 300 ternary graphs with the highest CH index values were selected, from which the graphs were picked out. The results reveal that there is no suitable diagram to distinguish between mineralization and barren granites. Further, in the diagrams to distinguish the type of mineralization, there are indeed some overlapping parameter ranges (such as Gd/Yb and Ce/Gd); however, it still can be observed that some binary and ternary graphs containing Eu or Ce (e.g. Eu/Gd, Eu/Tb and Ce/Sm) can discriminate between samples related to porphyry Cu and samples related to W–Sn (Sn). Some diagrams (i.e. Dy/Lu–Er/Lu, Gd/Dy–Er/Yb) also have been observed to distinguish between ore-forming types of granites, the underlying geochemical mechanism of which has not been fully understood. In addition, our results suggest that the linear discrimination analysis (LDA) method outperforms any of the traditional discrimination diagrams in discriminating ore-bearing and fertile granites. The results of geochemical data mining in this paper can be used as prospective indicators to provide a scientific basis for the study of Yanshanian hydrothermal deposits in South China. Moreover, the compositions of zircon are relatively simple to obtain, and the LDA methods described in this study could potentially be used as a useful exploration tool.

中文翻译：

---

通过锆石 REE 成分对比花岗岩成矿：从数据挖掘的角度

摘要 华南钦州—杭州湾燕山期岩浆活动发育良好，具有明显的成矿特异性。花岗岩中的锆石在成矿特异性研究中具有巨大的潜力。此外，随着原位锆石分析技术的发展，近年来获得了大量有关锆石成分的数据。因此，为了进一步阐明锆石特征在成矿富集性方面的独特诊断特征以及对成矿类型花岗岩的偏好，本文使用一组 13 种稀土元素以及燕山期已发表的锆石数据，详尽地探讨了所有可能的二元和三元判别图。 ，华南。这导致 4095 个二元图和 12,485 个三元图，根据他们对训练数据进行分类的成功程度进行排名。选取CH指标值最高的100张二元图和300张三元图，从中挑出图。结果表明，没有合适的图表来区分矿化和贫瘠的花岗岩。此外，在区分矿化类型的图表中，确实存在一些重叠的参数范围（例如Gd/Yb和Ce/Gd）；然而，仍然可以观察到一些包含 Eu 或 Ce 的二元和三元图（例如 Eu/Gd、Eu/Tb 和 Ce/Sm）可以区分与斑岩 Cu 相关的样品和与 W-Sn (Sn) 相关的样品。还观察到一些图表（即 Dy/Lu-Er/Lu、Gd/Dy-Er/Yb）来区分花岗岩的成矿类型，其潜在的地球化学机制尚未完全了解。此外，我们的结果表明，线性判别分析 (LDA) 方法在判别含矿花岗岩和肥沃花岗岩方面优于任何传统判别图。本文地球化学数据挖掘结果可作为前瞻性指标，为华南燕山期热液矿床研究提供科学依据。此外，锆石的成分相对容易获得，本研究中描述的 LDA 方法有可能用作有用的勘探工具。本文地球化学数据挖掘结果可作为前瞻性指标，为华南燕山期热液矿床研究提供科学依据。此外，锆石的成分相对容易获得，本研究中描述的 LDA 方法有可能用作有用的勘探工具。本文地球化学数据挖掘结果可作为前瞻性指标，为华南燕山期热液矿床研究提供科学依据。此外，锆石的成分相对容易获得，本研究中描述的 LDA 方法有可能用作有用的勘探工具。