Linear Regression Learning System 📈

An interactive, educational web application designed to help students visually and mathematically understand the foundations of Simple and Multiple Linear Regression. Built entirely in Python using Streamlit, this tool bridges the gap between raw code and theoretical mathematics by providing a hands-on, step-by-step learning environment.

🌐 Live Demo

You can try out the application live here: https://my-linear-regression-app.streamlit.app/

---

🎯 Educational Objectives

This system was developed with a strict focus on educational outcomes. It allows users to track the entire machine learning lifecycle without treating algorithms like a "black box."

1. Dataset Upload & Preview

Students can upload custom datasets (.csv) to experiment with data they care about. The system dynamically parses columns, differentiating between categorical and numerical features.

2. Preprocessing & Data Cleaning

A guided interface teaches students the importance of data hygiene.

• Missing Value Handling: Choose between mean imputation or dropping NaN rows.
• Categorical Encoding: Automatically encodes textual categories to numerical labels.
• Feature Scaling: Demonstrates Standardization (Z-Score) and Normalization (Min-Max) to explain why scaling is necessary for gradient descent convergence.

3. Exploratory Data Analysis (EDA)

Before training, students must analyze relationships in their data to ensure linear regression is an appropriate algorithm:

• Feature Distributions: Interactive Histograms and Box Plots.
• Correlation Analysis: A full Correlation Heatmap to identify collinearity and strong predictors.
• Relationship Visualization: Feature vs. Target scatter plots with OLS trendlines.

4. Interactive Learning Module (Theoretical Foundations)

A dedicated module that walks students through the math using proper $\LaTeX$ notation:

• Differentiates the formulas for Simple vs. Multiple Linear Regression.
• Defines Hypothesis Formulation ($h_\theta(x)$) and the Cost Function ($J(\theta)$).
• Step-by-Step Computation: Explicitly breaks down the math for the first sample of the dataset, showing the exact formula used, intermediate weight values, prediction calculation, and error calculation.
• Defines the Gradient Descent update rule mathematically.

5. Training Configuration (Gradient Descent)

Students can visually see the training process rather than just calling .fit():

• Manually configure Hyperparameters: Learning Rate ($\alpha$) and Epochs.
• Split data into Training and Testing sets.
• A live Cost Convergence Graph plots the Cost Function $J(\theta)$ over time, allowing students to visually understand convergence, underfitting, and over-shooting (exploding gradients).

6. Prediction & Evaluation

Evaluates the custom-trained model:

• Displays learned parameters (Weights and Bias).
• Reports Mean Squared Error (MSE), Mean Absolute Error (MAE), and R² Score.
• Allows users to input custom values into their trained model for live predictions.

---

🛠 Technical Architecture

• Frontend / Framework: Streamlit with heavy custom CSS injection to create a modern, dark-themed, React-like UI.
• Data Manipulation: pandas and numpy.
• Machine Learning: Custom Gradient Descent implementation built from scratch using numpy (No Scikit-Learn .fit() shortcuts were used for training to ensure educational transparency).
• Visualizations: plotly.express for interactive charts.

---

🚀 Installation & Execution

Prerequisites

• Python 3.8+ installed on your system.

Setup Instructions

1. Clone or Extract the Repository: Navigate to the project directory:

   cd Linear_Regression2

2. Create a Virtual Environment: It is recommended to use a virtual environment to manage dependencies.

   python -m venv venv

3. Activate the Virtual Environment:

   • On Windows:

     .\venv\Scripts\activate

   • On macOS/Linux:

     source venv/bin/activate

4. Install Dependencies: Install the required packages from requirements.txt:

   pip install -r requirements.txt

5. Run the Application: Launch the Streamlit server:

   streamlit run app.py

6. View the App: Your default web browser should automatically open the app. If not, navigate to http://localhost:8501.

---

Developed for interactive learning and academic demonstration.