AI-driven models for accountability and transparency in public sector administration

Accountability and transparency are crucial in today&s;s public administration, but as evidenced in traditional systems, inefficiencies, lack of access to data, and reduced participation of citizens are prevalent. This paper explores a framework using Python-based Artificial Intelligence (AI) designed to improve accountability and transparency in public administration. We develop a framework that uses both supervised and unsupervised machine learning models, Natural Language Processing (NLP), explainable AI (XAI), and visual analytics to extract, analyze, and visualize administrative data. The framework will identify anomalies, predict inefficiencies, and produce outputs that are interpretable for stakeholders. Evaluations based on publicly available datasets are shown to illustrate that Gradient Boosting produced high classification accuracy; K-Means effectively clustered administrative processes and citizen feedback; and explainability tools like SHAP and LIME were effective in producing transparent and interpretable automated decisions. Findings indicate that AI driven solutions can increase governance efficiency, increase public trust, and facilitate evidence-based decision-making, and move toward a more accountable and participatory public administration.