Machine Learning-Based Analysis of Research Policy Impacts on Academic Performance Metrics

Abstract

Artificial Intelligence (AI) has significantly transformed research policy analysis by enabling data-driven decision-making and predictive analytics. Research policies play a crucial role in shaping academic and institutional research outcomes, influencing funding allocation, citation impact, and overall research productivity. This study employs Machine Learning (ML) techniques to examine the effect of policy, outcome, and control factors on research performance. The dataset, sourced from an online repository on GitHub, underwent preprocessing steps, including handling missing values and feature selection to identify the most influential variables. The study applies Linear Support Vector Classifier (LinearSVC) to predict three key research metrics: research productivity, citation impact, and innovation index. Experimental results demonstrate that LinearSVC achieved an accuracy of 97% (F1-score: 0.96, AUC-ROC: 0.62) for research productivity prediction, 94% (F1-score: 0.93, AUC-ROC: 0.93) for citation impact, and 95% (F1-score: 0.95, AUC-ROC: 0.96) for the innovation index. These findings highlight the robustness of AI-driven analysis in evaluating research policies and optimizing institutional strategies. By leveraging ML models, this study provides insights into policy effectiveness, enabling institutions to refine their research strategies for improved scientific innovation and societal advancement.