Inspiration
We noticed an important difference in the amount of use health insurance was getting in the wake of COVID so we wanted to analyze if health insurance had an effect on types of vehicle insurance people with certain features and vehicles used.
What it does
Our model predicts if a person is interested in taking health insurance from the same company if they have vehicle insurance. It also predicts how the other features of vehicle insurance and person affect the decision for finalizing health insurance.
How we built it
We planned to use python because there are several inbuilt packages for developing such models. Second, we used R for plotting the graph because it is easier to do that in R Next was trying several models to find which model works the best for this type of dataset. We made a decision to not use the Neural Network because it is a simpler dataset and for a simple dataset, using a neural net will make it slow. At last, we made the model work faster by pre-processing the data. We normalized the data, removed redundancies, changed non-numerical values to numerical values, etc. Finally, we were able to achieve an accuracy of 87.8%.
Challenges we ran into
The first challenge was to find an appropriate dataset with many relevant features and values Trying out a good estimator was another challenge that we faced. We used several algorithms to predict but in the end we found the XGBoost model to be the best for this dataset. Next, our model was pretty slow, to make it more accurate and fast, we normalized the features, found the correlation matrix, and removed the redundant features. This made our model learn about 30% faster.
Accomplishments that we're proud of
We are proud of finding a correlation between vehicle insurance and health insurance because this would allow us to predict which insurance methods are more popular.
What we learned
We learned that while finding data alone is very challenging, the even more difficult part is finding data that will give us signals for our hypothesis test.
What's next for Correlation Between Vehicle and Health Insurance
There are still many challenges that can be accomplished like adding the COVID, salary of the household, type of car owned, a place where the person is staying. These can be very important features for the prediction. Next, using this prediction, the insurance company can devise some schemes or provide offers/ discounts to the clients to build client trust in them.
Log in or sign up for Devpost to join the conversation.