Enhancing the Efficiency of National R&D Programs Using Machine Learning-Based Anomaly Detection

This study is grounded on the premise that, given the transformative advances in artificial intelligence (AI) technologies occurring across the industrial landscape, AI tools should be actively implemented into the design and implementation of industrial policy. We argue that this is especially true for R&D policy, which is central to national competitiveness in science and technology, and which must consider multiple diverse variables, including the global economy, the overall industrial environment, corporate management, and technological capabilities. For this study, I apply machine learning (ML)-based anomaly detection (AD) to analyze high-performing national R&D projects, and specifically assess ML-based AD that considers both input and output variables and analyzes structural patterns. Building on these analytical results, I propose firm-size-specific differentiated policy measures designed to enhance R&D performance. The goal of this study is to establish a policy-decision framework that improves timeliness and precision in the operation and management of national R&D programs and, in the longer term, contributes to the realization of AI-based policy planning and operational management.