- Home
- Archives
- 2026
- Vol. 5 Issue 3
- Article
AI-Based Unified Government Scheme Recommendation: A Hybrid Rule-Filtered and Learning-Ranked Prototype for Citizen Scheme Discovery
Authors
1 2 3 4 5 Department of Computer Science and Engineering (Artificial Intelligence and Machine Learning), AMC Engineering College, Bannerghatta Road, Bengaluru, Karnataka, India.
Article Information
- Published In
- Volume 5, Issue 3 (2026)
- Published Online
- September-December 2026
- Pages
- 965-975
- Article Type
- Original Article
Abstract
View PDFWelfare programmes announced by central and state governments in India are numerous, but a substantial share of the population that qualifies for them never claims the benefit. The reasons are practical rather than legal: scheme information sits on many different departmental websites, eligibility rules are written in administrative language, and a citizen has no easy way of establishing which of the available programmes actually apply to their own circumstances. This paper reports the design and the current prototype implementation of GovScheme AI, a unified web platform that converts a short structured citizen profile into an ordered list of schemes the citizen appears to qualify for, together with a direct link to the corresponding official government portal. The platform gathers eight declared attributes — age, gender, highest educational qualification, present occupation, social category, annual family income, state of residence, and district — and passes them through a two-stage pipeline: a deterministic rule filter that discards schemes whose documented eligibility conditions the profile fails, followed by a relevance-scoring stage implemented with the scikit-learn library that orders the surviving schemes before they are displayed. At the time of writing, the catalogue, the profile capture form, the filtered recommendation view, the citizen dashboard and the outbound links to verified government portals are operational over a curated set of thirteen national employment and training schemes; the learning-based ranking stage and the feedback-driven model refresh remain at an early stage. No controlled experiment has yet been carried out, and consequently this paper reports no accuracy, precision, recall or user-study figures. Its contribution is a documented architecture, a working end-to-end prototype, an explicit account of what is and is not implemented, and a concrete evaluation protocol that subsequent work can follow.
Related Articles
Share Article
*Instagram doesn't support direct link sharing from web. Copy the link and share it in your Instagram story or post.