<![CDATA[Purpose - This study reviews recent research on clustering mixed-type data and examines how clustering algorithms, distance or similarity measures, validation methods, and application domains are combined, with particular attention to education and special education. Design/methods/approach - A systematic literature review with descriptive evidence mapping was conducted using Scopus-indexed studies published from 2020 to 2025. The search identified 2,065 records, and the documented selection process resulted in 57 included studies. Each study was mapped by clustering algorithm, distance or similarity function, internal validation, external validation, and application domain. Findings - K-Means was the most frequently reported algorithm (18 studies), followed by HDBSCAN (10) and DBSCAN (9). Euclidean distance appeared in 49 studies, while Gower distance appeared in one. Internal validation was not reported in 35 studies and external validation was not reported in 36. When validation was reported, the Silhouette Score and Accuracy were the most common measures, while DBCV and Adjusted Rand Index (ARI) were uncommon. Research implications/limitations - The findings show a recurring gap between heterogeneous data structures and methods that are mainly designed for numerical or compact cluster structures. Future studies should test mixed-type distance measures, density-based clustering, and compatible internal and external validation. The review is limited to Scopus, English-language publications, the 2020-2025 period, accessible full text under the review protocol, and the information available in the review record. Originality/value - The review connects similarity representation, clustering structure, and validation instead of discussing each component separately. It identifies a weighted Gower-HDBSCAN-DBCV-ARI configuration as a testable research direction. This configuration is not presented as an empirically validated or superior method.]]>
