<![CDATA[The heterogeneity of micro, small, and medium enterprises (MSMEs) limits the usefulness of uniform development programmes. This study develops a transparent financial-capacity segmentation of registered MSMEs in Bekasi City using the Knowledge Discovery in Databases framework. The administrative source contained 7,964 records and 15 fields for 2019–2024. After data-quality screening and exclusion of 41 records with unusable capital or turnover values, 7,923 records were analysed. Business capital and turnover were selected as the two K-Means inputs because they directly represent financial capacity and were sufficiently complete for all retained observations. District, a nominal variable, was excluded from Euclidean-distance calculation and used only for post-hoc geographic profiling. The revised implementation specifies k-means++ initialisation, random_state = 42, n_init = 50, max_iter = 300, and tolerance = 10−4. A common candidate range of K = 2–10 was evaluated using inertia and the Davies–Bouldin Index (DBI). Both diagnostics supported K = 3; the minimum DBI was 0.5940. The clusters comprised 7,789 (98.31%), 22 (0.28%), and 112 (1.41%) businesses, representing low capital–low turnover, high capital–very high turnover, and high capital–moderate turnover profiles. The contribution is not a new clustering algorithm, but a reproducible city-scale evidence pipeline that corrects nominal-feature handling, reports the preprocessing audit, and converts financial profiles into testable programme hypotheses. Because the two minority clusters may represent financial extremes, cluster membership should be verified with business sector, age, employment, and stakeholder evidence before policy use.]]>
Copyrights © 2026
