Heroku Files for Vehicle Sales Price Reccomender App

This repository is for hosting the files used to build the Heroku web app based on the analysis and models built on a vehicle sales dataset. An image of the front-end is displayed in Figure 1. Below you will find an overview of the code and files.

Figure 1: Front end of Heroku app using the SVR model.

Code Used

Python Version: 3.8.11
Packages: pandas, numpy, scipy, sklearn, flask, pickle
For Web Framework Requirements: pip install -r requirements.txt

Files

templates/index.html

This file contains the HTML code used for the front end of the app hosted with Heroku.

Procfile

This file contains the commands for running the application's containers. It specifies the application server as Gunicorn.

app.py

This file contains the flask actions and input manipulation performed for the application.

scaler.pkl

This file is the pickle containing the MinMaxScaler after it had been fit to the training data so that it can transform the user input.

support_vector_regression_model.pkl

This file is the pickle containing the hyperparameter tuned SVR after being trained on the training data so that it can be applied to the user input.

transform_for_prediction.py

This file contains a function used in the app.py file to process the user input (i.e., create dummy variables and scale) so that the model can make a prediction.

Resources

1. Kaggle: Vehicle dataset
2. YouTube: Live- Implementation of End To End Kaggle Machine Learning Project With Deployment by Krish Naik
3. IMAGECOLORPICKER.com