X-ray diffraction (XRD) is widely used to determine mineral compositions for depositional environment analysis.However, this composition analysis is time-intensive and subject to expert interpretation.To address these limitations, machine learning-based composition analysis models have been developed, but their reliability decreased for samples with unusual compositions, necessitating expert analysis.This study proposes a k-means clustering model to classify 488 XRD data into expert and machine learning groups based solely on intensity profiles.To enhance clustering performance, the optimal number of clusters (3-7) was evaluated using false positive (FP) and false negative (FN).The five-cluster model minimized both errors, assigning unusual samples to the expert group (low FP) and usual samples to the machine learning group (low FN).Applying the previously developed mineral composition analysis model to 38 usual samples reduced the mean and standard deviation of mean absolute error by 24% and 60%, respectively, compared to all 49 test samples.For 11 expert group samples, the clustering model provided information on minerals with unusual compositions, aiding expert analysis.Furthermore, validation using XRD data from the Korea Plateau demonstrated the model's scalability.The proposed method enhances the reliability of machine learning-based composition analysis model while providing valuable guidance for expert analysis.
Park et al. (Mon,) studied this question.