Similarity Identification System Design for Engineer to Order Products with SVD Matrix Factorization

ETO products, characterized by high variation and low quantity, present unique manufacturing challenges. These products are custom designed to meet specific customer needs, leading to a wide range of configurations and features. This customization involves complex raw material processes and inventory management, requiring meticulous documentation of shared materials across products. This research addresses the issue through an empirical analysis within a heavy manufacturing company, proposing a system to identify product similarities based on parameters such as material structure and specific product features. The system incorporates the Singular Value Decomposition (SVD) matrix method to systematically identify product patterns and characteristics. Codification mechanisms based on product feature parameters and matrix matriculation have been developed to identify new products through similarity calculations. The system generates a product network map based on these similarity values, supporting the documentation of shared materials in inventory management. This methodology constructs a similarity matrix using the Frobenius norm and cosine similarity of SVD components from product parameters, revealing the closeness of ETO products, providing insights for manufacturing preparation and potentially improve efficiency in relevant business process activity.