📃Project overview

Often times when we face any kind of symptom, we don't know what kind of preliminary disease it could be. Several diseases are associated with union of symptoms, thus hard to take any possible action over it. We in a team of 3, developed a machine learning web application as a part of our final year undergraduate project in a team of 3, that acts as a personal health assistance on several levels. The project resulted into a publication at IJRASET.

January 2019 – April 2019

Tech 👨🏻‍💻

Python, Numpy, Pandas, Scipy, Django, GPS, HTML, CSS

📕Source code:

Github

🗯️ The Problem

Many times while feeling unwell, we know the symptoms but don't know what exactly we are suffering from. So just by knowing the diseases we get an idea what could all we be suffering from and take necessary action as soon as possible. To solve this problem, we developed a web application that intakes the symptoms given by the user and returns probable diseases based on a trained ML algorithm.

🎯 Architecture

This web application was built upon client server architecture, where Django was used at the server side and traditional HTML, CSS on client side. The client would enter/ select the symptoms of afflicted body part, the result of which would be given by trained machine learning algorithm.

Process:

Following steps were performed over the completion of project:

• Initially we collected the data pertaining to the symptoms-disease relationship from here.
• The raw dataset was then pre-processed and cleaned using Numpy, Pandas libraries available in Python.
• The clean dataset was then converted into the dataset that could be used by Random Forest algorithm and be trained to be used at the server side. The training resulted into an accuracy of over 90%.
• This trained model was then incorporated into a GUI model, with backend endpoints being created using Django framework.
• The front end of the web application was developed using HTML, CSS and Bootstrap.
• The server on receiving the input symptoms then returns a JSON of all probable diseases, using the trained model the user could be suffering with along with their probable percentages.
• Further, the health assistance also suggests the nearby physicians available around the patient using GeoPositioning system.
• Additionally the user could also use the price comparator to get the minimal and equally effective generic medicines at a certainly lower prices.
• The project resulted into getting published at International Journal for Research in Applied Science and Engineering Technology (IJRASET). It can be checked at this link.

Some Visuals from various states of the application:

                                                                                                                                                          Made with 🖤 by Gulshan