Validate Your ML Model Performance on Term Deposit Dataset from Kaggle

This repository contains all the files for a ML prediction-checking web app.

Actual solutions are shielded from the public using Streamlit's secrets management.

Try it now:

Credit to:

1. Brajesh Mohapatra for the dataset hosted on Kaggle.

2. The Streamlit team for the Streamlit framework and for making Python deployment easier.

How to use the web app

1. Download the Term Deposit Prediction Dataset from Kaggle. In the dataset, there are several files:

   • train.csv: A CSV dataset containing the instances for training your ML model (with labels).
   • test.csv: A CSV dataset containing the instances for testing your ML model (without labels).
   • solution_checker.xlsx: A Excel file for you to check the performance of your ML model locally.

2. Train your model using train.csv.

3. Predict with your model using test.csv. Your output should be a .csv file with no column headers, and contains 13,564 instances.

4. Go to the web app and upload your prediction .csv file at the sidebar.

5. Check your prediction results - accuracy, precision, recall and F1-score.

6. Share your work with the world!

How to use this repository

1. You may choose to use Anaconda or Miniconda for environment setup. I used Miniconda for this project.

2. Clone this project repo into your local.

3. Before you can use this for your own project, you need to perform some manual steps.

   • You need to create a test folder to store your example solution for testing the web app.
   • You need to create a .streamlit folder under webapp folder to store your secrets (aka your answers).
   • Your answers should be in a file named secrets.toml so that you can use Streamlit's secret management.

   

4. Now, your local repo should be ready for further modification.

5. To test, go into webapp directory and run Streamlit:

   cd webapp
   streamlit run streamlit_app.py
   

For detailed documentation, you may refer to my Medium article here.

Happy tweaking!

---

END

---