Backorder Prediction - Django Web Application

This is a Django web application that uses machine learning to predict whether a product will go on backorder or not. It uses a pre-trained Random Forest Classifier, Decision Tree and LGBM models to make predictions based on various features such as product availability, lead time, and more.

Table of Contents

• Installation and Usage
• Features
• Development
• License
• Credits

Installation and Usage

Cloning the Project

To clone this project, run the following command in your terminal:

git clone https://github.com/Pradyothsp/backorder-prediction.git

Creating Virtual Environment

python3 -m venv venv
source venv/bin/activate

Getting Started

Navigate to the project directory and follow the steps below to set up and run the project locally:

cd app

1. Install the required dependencies by running the following command:

   pip install -r requirements/local.txt

2. Run the Django migrations to set up the database:

   python3 manage.py makemigrations
   python3 manage.py migrate

3. Start the Django development server:

   python3 manage.py runserver

4. Open your web browser and navigate to http://localhost:8000/ to view the application.

   The application has a simple web interface where you can input the product features and get a prediction on whether the product will go on backorder or not.

   You can also make predictions using the API endpoint by sending a POST request to http://localhost:8000/predict/ with a JSON payload containing the product features.

Features

• Backorder Prediction: Predict whether a product will go on backorder or not based on various features.
• Batch Prediction: Provide a CSV file containing multiple product records for batch prediction of backorder status.
• Single Product Prediction: Offer a form interface to input the features of a single product and obtain a prediction on its backorder status.

Result

The table below shows the performance metrics of different models on the backorder prediction task. The models evaluated are Decision Tree, Random Forest, and Light GBM. The dataset used for evaluation consists of train, validation, and test sets.

Model	Data Set	Accuracy	Recall	Precision
Decision Tree	Train	0.946	0.9651	0.9814
	Valid	0.8818	0.9079	0.8445
	Test	0.8633	0.8065	0.0605
Random Forest	Train	0.9981	0.9982	0.9998
	Valid	0.9194	0.9513	0.9727
	Test	0.9003	0.8077	0.1878
Light GBM	Train	0.9941	0.9954	0.9997
	Valid	0.9221	0.9513	0.9675
	Test	0.9069	0.7939	0.2052

Screenshots

Home Page:

Predict:

Result (Single Product Prediction):

Result (Batch Prediction):

Development

This application was developed using Django 4.2. The backorder prediction model was trained using scikit-learn and is stored in the backorder/model/backorder_best_model.pkl file.

To train a new prediction model, you can run the model.ipynb Jupyter Notebook in the root directory.

License

Backorder Prediction is licensed under the Apache License 2.0. See the LICENSE file for details.

The Apache License 2.0 is a permissive open source license that grants permissions to use, copy, modify, and distribute the software. It includes limitations on liability and requires that any modified or redistributed versions of the software be accompanied by a prominent notice stating the changes made.

You can find more information about the Apache License 2.0 here.

Credits

This project was created by Pradyoth S P.

The backorder dataset used to train the prediction model is from the Kaggle.