Web-based recommendation system for healthcare services

Chew, Jing Jia (2026) Web-based recommendation system for healthcare services. Final Year Project, UTAR.

PDF Download (8Mb)

Abstract

As the healthcare industry undergoes rapid digital transformation, patients are increasingly overwhelmed, where navigating thousands of unstructured online reviews to find specialised medical providers results in decision fatigue. This project aims to develop an intelligent web-based recommendation system that utilises machine learning to deliver personalised, localised healthcare suggestions tailored to individual patient symptoms and geographical context. The research methodology involved the collection and preprocessing of a dataset comprising 10,140 reviews from 15 major public and private hospitals in Penang, extracted using the Phantom GMB Audit Tool. The framework implements a hybrid machine learning architecture that integrates SVD for collaborative filtering with SVM for content-based sentiment classification, further enhanced by TF-IDF vectorisation and VADER sentiment analysis to interpret the semantic nuances of patient feedback. The system was developed using the Flask web framework and HTML to provide a seamless interface between the complex backend algorithms and the end user. Experimental results indicated that the hybrid model significantly outperforms standalone collaborative filtering models, achieving an accuracy of 80.10%, an RMSE of 0.6313, and an MAE of 0.3137. The project successfully designed a user-friendly interface with features, including a multilingual toggle (English and Chinese) to support Malaysia’s multiracial society, real-time location-aware prioritisation using the Haversine distance formula, and a specialised module that provides transparency into the ranking logic through keyword-matching insights. This study is highly significant as it transforms unstructured patient feedback into a structured, accessible decision-support tool that empowers the community to make more informed medical choices.

Actions (login required)