An advanced hospital rating system using machine learning and natural language processing

Abstract

Bangladesh is the host of 255 public and 4,452 private hospitals. Unfortunately, there is no reliable metric or resource available online to determine which hospital is better. Patients and their peers often find it difficult to choose the best hospital for their medical attention. The traditional star based rating system can easily be manipulated and they do not take user reviews into count. This is Where this Research and its techniques become useful. Our advanced hospital rating system takes reviews of a hospital and rates it based on the sentiment of the reviews. Our proposed model uses NLP and ML to rate the hospital solely based on the experience of the user shared online. That is why it not only rates the hospital but also identifies the strength and weaknesses of the institution. For this research, 14,443 unstructured reviews were collected from Google Maps of the top 38 hospitals in Dhaka. Additionally, these hospitals were rated based on their review’s sentiment and ranked according to the positive percentage. Basically, two types of ranking were introduced where in the general ranking system IBN Sina Specialized Hospital secures the first position and in Class based ranking system Square Hospital secures the first position. Furthermore, a web service is proposed where this trained model predicts the sentiment of the user’s reviews and ranks that institution. For future prediction, these reviews were created into multiclass datasets and pre-processed using NLP techniques, and trained into four machine learning models and two deep learning models to predict the sentiment. The most promising model is the Support Vector Machine (SVM) with an accuracy of 85.32%. it’s Precision, Recall and F1- score is 86%, 85% and 77% respectively.