Evaluation of computational techniques for purchase recommendation systems in a long steel industry

"The steel industry faces increasing challenges in managing product diversity, maintaining competitiveness, and improving operational efficiency. Within this context, recommendation systems present a promising opportunity to optimize sales processes, enhance customer experience, and support inventory management strategies. This study investigates the application of computational recommendation techniques to the long steel segment, following a structured methodology encompassing data collection from sales records, customer profiles, and product attributes, followed by exploratory data analysis to identify key patterns and correlations. Subsequently, multiple recommendation algorithms are developed and evaluated, including content-based filtering and collaborative filtering methods. Performance is assessed using precision, recall, F1-score, novelty, and RMSE metrics. The results offer insights into the adaptation of recommendation models for industrial B2B sales environments, highlighting their potential to boost sales performance, increase customer engagement, and improve supply chain responsiveness. Furthermore, the study discusses specific challenges encountered, such as data sparsity, cold-start issues, and the critical role of domain-specific feature engineering. By addressing these challenges and leveraging advanced machine learning techniques, this research lays a foundation for future initiatives aimed at AI-driven sales optimization in the steel industry."