Application of the DBSCAN Algorithm in MSME Clustering using the Silhouette Coefficient Method

MSMEs participate in the very important contribution of developing Indonesia's economy, where this industry contributes to GDP and also to the absorption of labor. Most MSMEs in Sidoarjo Regency are still constrained by financial management and the utilization of technology. This research will apply the DBSCAN method to clustering MSMEs in Sidoarjo for the purpose of finding patterns in characteristics related to capital, turnover, and workforce. The analysis will involve 1,479 MSMEs, while the research methodology applies the CRISP-DM method to guide the process from business understanding up to the implementation phase. Normalization using Simple Feature Scaling was applied before clustering. The results of this analysis provide insight that the best possible combination of the parameters in DBSCAN is epsilon (ε) 0.10 and MinPts 16, which gives the optimal value of Silhouette Score as 0.4304. It creates seven clusters, in which the third has the highest Silhouette value of 0.9326, indicating that there are high similarities recorded within that cluster. These results provide essential lessons to develop more targeted policy strategies and interventions for MSMEs in Sidoarjo and explore the capabilities of DBSCAN as an effective analytical tool in determining the characteristics of businesses in the region.