Application of Deep Learning Algorithms in Clustering Production Units: Enhancing Regulatory Processes Using Soft Computing Approaches (Case Study: North Khorasan Province)

The oversight of production units ensures compliance with national and international standards. The National Standards Organization gathers the quality documentation of manufacturers within the SINA system. Despite this extensive database, the intensity and frequency of sampling and inspection processes are conducted without considering quality records. This study develops an intelligent clustering and deep learning approach to categorize production units in North Khorasan. In this regard, algorithms such as Spectral clustering, Density-based hierarchical clustering with noise detection, Deep embedded clustering, Gaussian mixture model, and Louvain were implemented. The Davies–Bouldin index, Average Silhouette score, and Calinski–Harabasz index were utilized to assess the quality of clusters. Furthermore, the efficacy of the chosen approach was contrasted with the findings of earlier studies (K-means). The results showed that Deep embedded clustering outperformed, revealing hidden data structures with more cohesive, separable clusters and superior metrics. Deep embedded clustering outperformed K-means, reducing Davies–Bouldin by ۰.۲۴, increasing silhouette by ۰.۰۵۲, and enhancing Calinski–Harabasz by ۴۷۹. It identified five distinct production clusters by location, production volume, and quality metrics, enabling more efficient monitoring.

کلیدواژه ها:

Spectral Clustering ، HDBSCAN ، Deep Embedded Clustering ، Gaussian mixture model ، Production units