This repository contains a web application for building and deploying a Random Forest model, with options for both regression and classification tasks. The app uses Python's Flask for the backend and HTML, CSS, and JavaScript for the frontend.
Users can upload a "cleaned" dataset in CSV or XLSX format. This dataset should contain no missing values or prior transformations. Once uploaded, the app generates a Cramér’s V heatmap to visualize correlations among categorical features.
Based on the heatmap, users can specify how to encode the categorical features. With encoding selected, the app trains a Random Forest model, splits the data, and displays performance metrics.
The trained model is saved and can make predictions on a single-row CSV file, excluding the target column.
First, clone this repo to your local machine. Then, create a new Conda environment with its dependencies by executing the following command in the terminal or console.
Note: This repo is accelerated with Intel(R) Extension for Scikit-learn. So, make sure to clone this repo on an Intel-powered local machine.
conda env create --file requirements.yml
After the environment is created, activate it with:
conda activate ml-any-table
To launch the web application, execute the following command in the terminal/console.
Ensure that your current working directory is set to the cloned repo.
python backend.py
Once the Debugger is activated! message appear in the terminal, open the browser (Chrome, Firefox, Safari, etc.) and go to localhost:8000. The web application is now ready to use!
The tutorial video on how to use the web app is provided below.