Health-Sleuth

home-page
provide the clues
conducts investigation - risk of hypertension
conducts investigation - healthy
conducts investigation - risk of diabetes

Inspiration

Diseases such as diabetes, depression and hypertension are bit hard to diagnose properly take the example of hypertension. Your blood pressure can rise due to anxiety when you visit the doctor - this condition is known as whitecoat hypertension and it increases the chances of misdiagnosis. On the other hand, some people have masked hypertension. Depression affects one in three people with hypertension while one in four patients with diabetes have depression. It is estimated that one out of three adults has hypertension and that more than 50% of them are unaware of this condition.

Our lifestyle has a lot of impact on these diseases. Baseline daily and non-daily smoking was associated with depression .Low physical activity and heavy alcohol drinking were associated with persistent depression.

We wanted to create a project to conduct a health-investigation and diagnose a person with high or low risk of these diseases based on their lifestyle parameters.

What it does

Health sleuth is essentially a disease-investigator which investigates about our health and risk to diabetes, depression and hypertension. It is a machine learning model which takes in some clues from the user to conduct the investigation and can predict whether a person has low or high risk of diabetes, depression and hypertension. Health Sleuth can also predicts whether a person is completely healthy or not.

How we built it

First we got our dataset then we pre-processed the data, split it and trained our model .Our model uses the algorithm: Random Forest. We specifically chose random forest algorithm for our model as it provided the highest model accuracy compared to the other algorithms we tried out. A random forest is a meta-estimator which aggregates many decision trees. It is one of the most accurate learning algorithms available. For many data sets, it produces a highly accurate classifier.For our dataset we were able to receive a model accuracy of .8719 . We used streamlit to build the front-end for the application.

Challenges we ran into

We initially had some trouble trying to find an apt dataset for our project but we finally got one after spending some considerable time and effort. We also had a problem of deciding the algorithm to be used as we tried random forest regressor, decision tree regressor and linear regressor but we finally chose random forest regressor as it has the highest model accuracy.

Accomplishments that we're proud of

In this short span of time we were able to build a successful project to detect low and high risk of diabetes, hypertension and depression. Also this was the very first time that we were trying out Machine Learning, which makes us super proud. We had multiple brainstorming sessions where we were able to come up with modifications to our base idea to help people out with these diseases right in their homes. The users can also tweak their lifestyle a bit from our model to see what predictions might come and decide on that basis. All these milestones made us very proud because we are progressing towards the greater good.

What we learned

Since this was the first time we were doing an ML project, we learned quite a lot. Thinking back to our working on this project, we learned how to build a machine learning model and finding the apt algorithm for it. While doing this project we got a deeper understanding of machine learning and how the model is able to perform the prediction.