An Intelligent Roommate Matching System Using Machine Learning for Smart Hostel Management

This paper presents an intelligent roommate matching system integrated into the HostelHaven Smart Hostel Management platform, designed to improve the quality of hostel life through data-driven decision-making. Traditional room allocation methods, such as random assignment or basic preference matching, often fail to consider deeper behavioral compatibility among students, leading to conflicts, discomfort, and reduced academic productivity.

The proposed system collects detailed lifestyle-based data, including sleeping patterns, cleanliness levels, study habits, noise tolerance, and social behavior. These attributes are preprocessed and transformed into numerical feature vectors suitable for machine learning analysis.

A hybrid methodology is adopted, where K-Means clustering groups students with similar behavioral characteristics, and cosine similarity is applied within each cluster to calculate fine-grained compatibility scores. These scores are normalized and presented as percentage values, making them easy to interpret for both administrators and students.

The system is implemented using a scalable full-stack architecture, consisting of React for the user interface, Node.js for backend services, MongoDB for efficient data storage, and a Flask-based microservice dedicated to machine learning computations. Furthermore, a manual override feature is incorporated, allowing administrators to make adjustments based on special cases or practical constraints.

By combining machine learning techniques with a user-friendly system design, the proposed solution enhances roommate compatibility, reduces interpersonal conflicts, and promotes a comfortable and productive living environment, ultimately improving student satisfaction and overall hostel management efficiency.