I have to buy health insurance now since I'm no longer buying the school's health insurance plan, but I have absolutely no idea how much I should be paying for it. I figured if I'm going through this problem, a lot of other people must be also going through the same thing. Then Justin suggested that people who already have health insurance can also benefit from the knowledge of knowing if they were overpaying for their plan. That's when we decided to build this project to help clueless people like me plus others who simply might be overpaying for something that is so essential.

What it does

The project takes some of your attributes as inputs and then uses machine learning to output what a predicted value of what you should actually be paying for your health insurance. So, suppose a person who is paying $10,000 per year for their health insurance enters their attributes onto our website will receive a prediction of what they should actually be paying. If the prediction is $8,000, then the person knows he's overpaying and needs to change his insurance provider. And for people who are going to be buying health insurance for the first time, they would know when an insurance provider is trying to charge them too much.

How we built it

We used a machine learning model in python to get the weights for our predictions. We uploaded the weights onto our JavaScript file as parameters. Then we used JavaScript and HTML to collect data from our users. Finally, we used JavaScript to predict costs and output it via HTML.

Challenges we ran into

We tried making the webpage look much nicer, but we quickly ran out of time. So, we sacrificed the beauty of the website in order to actually get it running. Then we ran into wrong predictions. Our model was outputting negative costs, which are impossible. We fixed that by debugging our code in Jfiddle. After that, we ran into a problem where our script file was running faster than our webpage and so we weren't getting our predictions as outputs. This took us the longest time to figure out how to solve. Finally, we were getting a low accuracy and so we had to train our model on different machine learning models to see which one worked the best.

Accomplishments that we're proud of

We made the website fully functional !! Although it looks ugly, it still outputs the correct results. We tested our model on our friends and we got the correct results every single time. We also figured out that one of our friends were overpaying. So yeah, we made something that is not only useful but also helpful for a lot of people.

What we learned

A lot about JavaScript. A lot about GitHub. A lot about machine learning models.

What's next for Health Insurance Price Predictor

We want to make the website more robust. Definitely prettier and more dynamic. We want to turn it into an app. We want the accuracy to be even higher and website to have more functionality. Finally, we want to make the machine learning model inside our JavaScript file.

Share this project: