TibebAI

Tibeb means wisdom in Amharic.
This repository contains machine learning experiments focused on predicting student performance using regression models.

Developed as part of the Elevvo Machine Learning Internship Program.

---

Project Overview

This project explores the relationship between study hours and exam scores using the Student Performance dataset.
The goal is to apply regression techniques to model, predict, and evaluate student outcomes.

---

Workflow

The notebook is organized into the following main steps:

1. Installing Dependencies

   • Required Python libraries for data analysis and machine learning.

2. Data Cleaning & Visualization

   • Download dataset from Kaggle using the API
   • Unzip and load data into Pandas
   • Inspect datatypes and missing values
   • Handle outliers (IQR method)
   • Visualize features to understand distributions

3. Splitting the Dataset

   • Train/test split to prepare for modeling

4. Training Linear Regression Model

   • Import Linear Regression from scikit-learn
   • Create model instance and fit with training data
   • Make predictions on test data
   • Evaluate model using metrics: MAE, RMSE, R²

5. Visualizing Results

   • Plot regression line
   • Scatter plot of actual vs predicted values

---

Tools & Libraries

• Python
• Pandas
• Matplotlib
• Scikit-learn
• Kaggle API

---

Covered Topics

• Regression (Linear & Polynomial)
• Model evaluation metrics (MAE, RMSE, R²)
• Data cleaning & handling outliers
• Data visualization

---

Repository Structure

TibebAI/
│── data/              # datasets (not uploaded if large)
│── notebooks/         # Google Colab / Jupyter notebooks
│── scripts/           # Python scripts for modular code
│── results/           # plots, model outputs, evaluation metrics
│── README.md          # project overview

---

Showcase

• GitHub Repository
• Google Colab Notebook Link

---

Example Results

• Predicted vs Actual Scores scatter plot
  

• Regression Line visualization
  

Future Work

• Experiment with polynomial regression for better fit
• Add more features (sleep, participation, parental support, etc.)
• Try advanced models: Random Forest, Gradient Boosting, Neural Networks
• Deploy model as a simple web app using Streamlit/Flask

---

Inspiration

This project is named TibebAI to highlight the value of wisdom (tibeb in Amharic) in both learning and technology.

---

Internship Info

This work was completed as part of the Elevvo Machine Learning Internship Program (August 2025).
The program emphasizes real-world projects, personalized feedback, and professional portfolio building.