Summative Assignment: Linear Regression and API Deployment

This repository contains all components of the summative assignment for BSE Machine Learning and Mathematics for Machine Learning. The project addresses a specific use case with linear regression and includes a comparison with other machine learning models, API deployment, and a Flutter-based mobile app for predictions.

My Mission and Use Case

Mission:

To empower young women in STEAM (Science, Technology, Engineering, Arts, and Mathematics) fields, foster DEI (Diversity, Equity, and Inclusion), and drive environmental sustainability for a better future.

Use Case:

The project aims to predict an individual's carbon footprint based on their daily habits, such as electronic usage, meal consumption, transportation, grocery expenditure, and clothing purchases. This use case is unique and specific, aligning with my mission to drive environmental sustainability for a better future.

Dataset

Description:

Dataset: The dataset contains variables representing daily habits and corresponding carbon footprint. Variables include:
- Electronic devices owned (integer)
- Screen time (hours/day) (float)
- Monthly grocery bill (float, in USD)
- Monthly transportation expenditure (float, in USD)
- New clothes bought monthly (integer)
- Meals consumed weekly (integer)
Source: Retrieved from Kaggle, titled Daily Habits and Carbon Footprint Dataset.
Dataset Link: https://www.kaggle.com/datasets/dumanmesut/individual-carbon-footprint-calculation

Key Visualizations:

Correlation Heatmap: Shows relationships among variables.
Variable Distribution: Histograms for each feature to understand spread and outliers.

Task Details

Task 1: Linear Regression Implementation

Implemented Linear Regression, Decision Trees, and Random Forest models.
Optimized Linear Regression using Gradient Descent.
Evaluated all models using metrics such as Mean Squared Error (MSE) and R² score.
Saved the best-performing model (Random Forest) as model.pkl.

Task 2: API Development

Developed a FastAPI-based prediction API with:
- POST endpoint: /predict
- Pydantic validation: Ensures proper input types and ranges.
- CORS middleware: For cross-origin access.
Hosted the API on Render with Swagger UI available at:
- Public Swagger URL

Task 3: Flutter Mobile App

Developed a mobile app with Flutter to interact with the API.
- Input Fields: Text boxes for all input variables.
- Button: "Predict" button to send data to the API.
- Output Display: Shows predicted carbon footprint or error messages.
The app fetches predictions using the deployed API.

Task 4: Video Demo

YouTube Demo: A 2-minute video showcasing:
- API functionality via Swagger UI.
- Mobile app making predictions using the API.
Video Link

How to Run the Project

Step 1: Linear Regression Model

Clone the repository.
Navigate to the summative/linear_regression/ directory.
Open multivariate.ipynb and run all cells in a Jupyter Notebook environment.

Step 2: API Deployment

Navigate to summative/API/ directory.
Install dependencies:
```
pip install -r requirements.txt
```
Run the API locally for testing:
```
uvicorn prediction:app --reload
```
Access Swagger UI at http://127.0.0.1:8000/docs or use the deployed API link.

Step 3: Flutter Mobile App

Navigate to summative/FlutterApp/.
Install dependencies:
```
flutter pub get
```
Run the app:
```
flutter run
```
Ensure the app connects to the API via the provided endpoint.

Evaluation Metrics

Linear Regression Model

Mean Squared Error (MSE): Evaluated on test data.
R² Score: Indicates goodness-of-fit.

API

Validated inputs using Pydantic models.
Implemented error handling for invalid input ranges.

Mobile App

Tested with multiple scenarios for prediction and error handling.
Ensured seamless communication with the deployed API.

Demo Links

Swagger UI: https://linear-regression-model-pjjl.onrender.com/docs
YouTube Video: https://youtu.be/0cbiCNLjREA

eadewusic/linear_regression_model