Building a Machine Learning Application with Django

This article outlines the process of creating an end-to-end machine learning application using the Django framework, specifically designed to serve predictions from a trained model. The tutorial focuses on utilizing the Iris dataset and demonstrates how to set up the project, train a model, and create a web interface for user interaction.

Building Machine Learning Application with Django

To begin, developers need to establish a project directory and install the necessary Python packages. The required packages include Django, scikit-learn, and joblib. The initial setup involves creating a new Django project named `mlapp` and an application called `predictor`. The project structure is organized to facilitate model training and web form creation.

Building Machine Learning Application with Django

The application utilizes the classic Iris dataset, which is readily available in scikit-learn. A script named `train.py` is created to load the dataset, split it into training and testing sets, and train a Random Forest classifier. Upon successful training, the model is saved in a specified directory using joblib, allowing it to be reused for predictions.

Building Machine Learning Application with Django

After setting up the model, developers must configure Django to recognize the new application. This includes registering the `predictor` app in the `INSTALLED_APPS` section of `settings.py` and ensuring that the templates directory is correctly set up for loading HTML files.

Building Machine Learning Application with Django

Django's URL routing is crucial for directing user requests to the appropriate views. The project-level URLs are defined in `mlapp/urls.py`, while app-specific routes are established in `predictor/urls.py`. This setup includes routes for the home page, prediction form, and API endpoints.

Building Machine Learning Application with Django

To facilitate user interaction, a form class is created using Django's built-in forms module. This form captures the necessary flower measurements (sepal and petal dimensions) required for the Iris classifier. The form is integrated into the web interface, allowing users to submit their data for predictions.

Building Machine Learning Application with Django

The application’s prediction logic is encapsulated in a dedicated `services.py` file within the `predictor` app. This file contains functions to load the trained model and make predictions based on user input. The predictions return both the predicted class and associated probabilities.

Building Machine Learning Application with Django

Three views are constructed to manage user interactions:

Building Machine Learning Application with Django

1. home: Renders the prediction form.

Building Machine Learning Application with Django

2. predict_view: Handles form submissions and displays results.

Building Machine Learning Application with Django

3. predict_api: Provides a JSON API endpoint for programmatic access to predictions.

Building Machine Learning Application with Django

To ensure the application functions correctly, Django's built-in testing framework is employed. Tests are created to verify that the homepage renders properly and that the API endpoint returns valid prediction responses. Running these tests confirms the application's reliability.

Building Machine Learning Application with Django

The tutorial successfully demonstrates how to build a complete machine learning application using Django, from model training to user interface creation. While this project centers on the Iris dataset, the foundational structure can be adapted for more complex models and larger datasets, paving the way for robust, production-ready machine learning applications.

Building Machine Learning Application with Django