A hybrid machine learning framework for optimized MSME classification and knowledge discovery in smart village systems

Micro, Small, and Medium Enterprises (MSMEs) are a vital component of local economic growth, making significant contributions to advancing Smart Village and Smart Economy programs. The classification and distribution processes to support MSMEs across numerous locations continue to be manual, subjective, and inefficient. This study employs a hybrid data mining methodology that combines the K-Means clustering algorithm with Random Forest classification to enhance the precision and impartiality of MSMEs categorization in Bogor Regency. The study adheres to the Knowledge Discovery in Databases (KDD) framework, which includes data selection, cleansing, transformation, mining, and evaluation. The datasets consist of 7,615 MSMEs obtained in early 2025 from the Bogor Regency Cooperative and MSMEs Office. Following preprocessing and feature selection, K-Means identified an optimal cluster count of k = 3 using the Elbow method. The clusters were used as training labels for the Random Forest model, which achieved 95.9% accuracy, 0.94 precision, and 0.94 recall. Cluster validation produced a Silhouette Score of 0.6926 and a Davies–Bouldin Index of 0.994, signifying robust separation and cohesion. The results indicate that the K-Means–Random Forest hybrid model markedly enhances classification efficacy and establishes a data-driven basis for empowering MSMEs with Smart Village and Smart Economy paradigms.