Car Price Prediction Model

• 

  Interface

Inspiration

I am interested in both AI and Data, having learned machine learning in Python recently. I have gotten my hands on building simple models using other tools but wanted to try using Python. I wanted to try doing regression models too as I have never developed them before. This problem statement was chosen as I am familiar with Carsome and wanted to have the chance of getting an internship or even further job opportunities in the future.

What it does

The model I made will predict the starting bid of cars sold in the Carsome platform through the given features in the dataset.

How we built it

I made a notebook to do most of the data mining process including simple data exploration, data visualization, pre-processing, model building, model evaluation, and optimization.The model deployment process involves creating a pickle file of the model, creating a simple html file as an interface to interact with the model, and another python file is then made to deploy the pickle model using the Python Flask Framework.

Challenges we ran into

One of the biggest challenges I have faced during the model-building process is the preprocessing stage, especially during encoding. As this is the first time I am doing a machine learning model using Scikit Learn, I am not really familiar with what encoding methods I have to use and how should I encode the dataset. Since this is the first time I got my hands on Flask, I am not quite familiar with the framework.

Accomplishments that we're proud of

I have successfully developed and deployed a regression model with good accuracy (~93%). I have also learned and applied much newly learned knowledge, especially for Python that I previously am not quite knowledgeable about.

What we learned

I learned new knowledge about machine learning and data mining processes in general such as pre-processing methods, how to handle datasets, etc. I have also learned new libraries and frameworks Scikit Learn and Flask.

What's next for Car Price Prediction Model

Further improvements on the implementation and interface. Better features can also be picked out to achieve a better-performing model in terms of accuracy and efficiency. Algorithms to continuously update frequency datasets for encoding purposes.

Built With

• html
• jupyter-notebook
• python
• scikit-learn