🚀 Project Overview
Traditional recommendation systems often rely on hard-coded logic that requires manual reprogramming to update. This project introduces an Agentic architecture that empowers academic experts to dynamically modify decision criteria and weights through a no-code interface.
🧠 Reasoning Core
-
The system utilizes a sophisticated reasoning engine powered by four parallel hybrid Multi-Criteria Decision-Making (MCDM) algorithms to process multidimensional student data:
-
q-ROF-AHP: Captures nuanced judgment and uncertainty through membership and hesitancy degrees.
-
BWM-VIKOR: Efficiently aggregates expert opinions to find compromise solutions among conflicting criteria.
-
SWARA-MOORA-3NAG: Employs multiple normalization techniques to ensure ranking robustness.
-
LTSF-CRITIC-EDAS: Provides objective, data-driven weighting combined with linguistic fuzzy logic.
📊 Key Features
-
Dual-Input Layer: Collects academic history (SPM and University grades), technical skills, and RIASEC personality traits from students while allowing weight configuration from advisors.
-
Psychological Alignment: Maps personality types (Investigative, Realistic, Artistic, etc.) to specific technical domains for better career congruence.
-
Top-2 Validation: Achieved high reliability in pilot testing, frequently matching expert judgments and identifying latent student aptitudes.