HackRU_Vitech_Challenge

• 

  Screenshot demonstrating 99.4% accuracy

Inspiration

One of the challenges of this hackathon included data analysis, and machine learning algorithms. These two topics were something unusual that we wanted to take as a challenge for our future careers as computer scientists.

What it does

This project includes visualizing and displaying useful data gathered on a website, as well as creating a machine learning algorithm that achieved ~98% of accuracy determining whether a previous dental insurance user is prospect to buying accident insurance.

How we built it

We began by downloading the data from the Vitech API using Python to a local mongodb database. We later exported it to a cvs file to graph the data using Tableau. Afterwards, we trained a deep neural network to classify participants based on whether or not they would buy an accidental insurance plan.

Challenges we ran into

Downloading data and retraining data was a slow process with multiple errors that got in our way to completion. Normalizing the data was also challenging as it was mainly trial and error. Also picking up the neural network depth and width was troublesome.

Accomplishments that we're proud of

We trained a model with 250,000 records and consistently achieved 99% accuracy using 15,000 test records.

What we learned

We expanded our knowledge on the programming languages JavaScript and Python, as well as Big Data. We learned how to use software like Tableau and mongodb for this project as well.

What's next for HackRU_Vitech_Challenge

Use real data to provide live data analysis and improving the model for real-life case scenarios.

Built With

• css
• html
• javascript
• python
• shell
• tableau
• tensorflow