E-Counsellor: An Intelligent Machine Learning-Based College Recommendation System for Maharashtra CAP Engineering Admissions

Abstract

The Centralized Admission Process (CAP) conducted by the Maharashtra State CET Cell allocates engineering seats to hundreds of thousands of aspirants annually. Students routinely struggle to interpret complex, unstructured cutoff PDFs spanning multiple rounds, categories, and institutions, resulting in suboptimal college choices and heightened decision stress. This paper presents E-Counsellor, an intelligent, data-driven college recommendation platform that automates the analysis of historical Maharashtra CAP cutoff data and applies an XGBoost regression model to predict personalised admission probabilities for each student-college-course-category combination. A Spring Boot REST backend orchestrates business logic, communicates with a PostgreSQL database, and delegates probability scoring to a FastAPI ML microservice via a batch inference endpoint, minimising inter-service latency. Student-facing results are ranked by admission probability and labelled with calibrated SAFE, MODERATE, or RISKY risk indicators. An Admin Panel secured by JWT-based authentication enables institutional data management. Evaluation confirms that E-Counsellor significantly reduces manual cutoff analysis effort and improves the quality and personalisation of college preference lists for engineering aspirants.