Car Price Prediction

Selling any used cars? Not sure if you're undervaluing it or overvaluing it? Don't fret, as my model will give you a great estimate to sell your car at!

Comment

Inspiration

I see people getting scammed whilst selling their cars, and so I believe people should be able to know what their car is worth with ease to promote fairness in the market.

What it does

If you have trouble knowing what your car is worth, you can input attributes of your car (model, mileage, year made) and it will give you a fairly accurate estimate of what your car should sell at

How we built it

  • I found a lengthy dataset on kaggle that fit this criteria and downloaded it as csv
  • pip installed torch, pandas and other libraries
  • Created a jupyter notebook and trained the model using Pytorch's Adam optimizer which adjusts the weights
  • Used matplot lib to visualize the actual vs predicted values as well as displaying the mse and r value.
  • Exported the model into my flask application where I then deployed it into a server
  • Used npm to create a Next JS application by Vercel, where I designed the frontend with libraries such as shadcn UI and recharts
  • Used an API to access the flask server to then send a request and query data
  • Used the recharts to design a nice chart for the user to visualize their requests.

Challenges we ran into

  • During my frontend development, while I was adding in user input of the model, I realized the model only works with labels it has seen before in the dataset.
  • I had to get a string array of all the unique possible models and add them into a dropdown menu
  • Using commands in the terminal wouldn't work for some reason and I had to manually add the directory to the PATH environment variable in order for jupyter and npm to work.

Accomplishments that we're proud of

  • Successfully communicating with the flask application (fullstack application)
  • Designing a clean and nice UI
  • Overall, being able to train the model into something that's fairly accurate given how challenging our problem is to solve

What we learned

  • We learned how it feels to be a fullstack developer, switching between frontend and backend rather than specializing in one field.
  • Also learned how to visualize actual vs predicted in a graph to show how accurate the model is.

What's next for Car Price Prediction

  • Hopefully expanding our model to include more cars
  • creating a DB and an authentication system using firebase or mySQL, maybe even a noSQL db such as mongoDB

https://www.canva.com/design/DAGbXdvyyPw/tiFGjkMTfIkAvxB7ymOfng/edit

Canva Presentation

Built With

Share this project:

Updates