Using the k-means method, it clusters 2965 character types classified in the JIS first level and detects similar Kanji characters. Kanji Clustering Analysis provides an API. Please see the documentation.