Iris Flower Classifier

This project is an implementation of Logistic Regression with Gradient Descent and the One-vs-Rest strategy for Multiclass Classification, built from scratch with NumPy and Pandas. It leverages the Iris Dataset from UCI Machine Learning repository to perform precise species classification based on the sepal and petal characteristics of iris flowers. The trained model's weights are then employed to serve predictions through a website powered by FastAPI and HTMX.

Preview

Live Version

This page is currently deployed. View the live website.

Implementation details

Features given in the dataset (input features):

• Petal length
• Petal width
• Sepal length
• Sepal width

Classes Classified (target labels):

• Iris-setosa
• Iris-versicolor
• Iris-virginica

Model used:

• m - training examples, w - weights vector, b - bias
• Logistic Regression Model: $$f_{w,b}(x) = g(\textbf{w} . \textbf{x} + b)$$ where g is the sigmoid function given by: $$g(z) = \frac{1}{e^{-z}}$$
• Cost function for logistic regression: $$\textbf{J}(\textbf{w}, b) = \frac{1}{m} \sum_{i = 0}^{m - 1} \left( loss(f_{\textbf{w}, b}(\textbf{x}^{(i)}), y^{(i)}) \right)$$ where the loss function is the cost for a single data point and is given by: $$loss(f_{\textbf{w},b}(x^{(i)}),y^{(i)}) = \left(−y^{(i)}\log(f_{\textbf{w},b}(\textbf{x}^{(i)})\right) − \left((1 − y^{(i)})\log(1 - f_{\textbf{w}, b}(\textbf{x}^{(i)}))\right)$$ where $f_{\textbf{w},b}(\textbf(x)^{(i)})$ is the model's prediction and $y^{(i)}$ is the actual label $$f_{\textbf{w},b}(\textbf{x}^{(i)}) = g(\textbf{w} . \textbf{x}^{(i)} + b)$$

Setup

• Clone this project:

  git clone https://github.com/Vyvy-vi/iris-flower-classifier/
  

• Install Python3
• Install dependencies

  pip install -r requirements.txt
  

• Run Jupyter Notebook

  jupyter notebook
  

• Run application

  python3 main.py
  

Usage

• Training the Model: To train the model and generate the weights and bias, run the classification-logistic-regression-from-scratch.ipynb Jupyter notebook. (run the jupyter notebook command)
• Running the Web Application: Execute the web application using python main.py. This starts the web server, making the prediction service available at http://localhost:8000.
• Making Predictions: Input sepal and petal measurements via the web interface and receive predictions for the iris flower species.

Feedback and Bugs

If you have feedback or a bug report, please feel free to open a GitHub issue!

License

This software is licensed under The MIT License.

Copyright 2023 Vyom Jain.