Skip to content

README.md

Latest commit

 

History

History
256 lines (178 loc) · 9.17 KB

README.md

File metadata and controls

256 lines (178 loc) · 9.17 KB

Student Performance Deep Learning

🎓 Student Performance Prediction Using Deep Learning

Deep Learning • Education Analytics • Data Visualization
Predicting student performance with TensorFlow neural networks and data-driven insights.

Visual summary of the project: predicting student performance using TensorFlow neural networks and educational analytics.

📑 Table of Contents

📘 Overview

This project applies Deep Learning to predict student performance from real-world data.
It covers the full machine learning lifecycle — data exploration, exploratory data analysis, preprocessing, modeling, and evaluation — to extract actionable insights that enhance learning outcomes.

🔍 Objectives

  • Analyze and visualize student performance patterns
  • Build predictive models using Artificial Neural Networks (ANNs)
  • Evaluate results using key metrics for regression and classification

⚡ Key Features

✅ Complete data science pipeline (EDA → Modeling → Evaluation)
✅ Built with TensorFlow/Keras for scalable deep learning
✅ Interactive Jupyter Notebook walkthrough
✅ Beautiful data visualizations for insights and interpretability
✅ Reproducible and modular — ideal for research or education projects

📂 Repository Structure

Path Description
.gitignore Git configuration to exclude unnecessary files
LICENSE MIT license for open-source distribution
README.md Project overview, methodology, and visualizations
requirements.txt Python dependencies for running the project
StudentPerformanceFactors.csv Dataset containing student performance features
student_performance_dl_analysis.ipynb Main notebook with EDA, preprocessing, modeling, and evaluation
assets/ Folder contains visual assets used in the README (plots, thumbnails, etc.)

🧪 Technologies Used

Category Tools / Libraries
Language Python 3.10+
Data Processing NumPy, Pandas, Scikit-learn
Visualization Matplotlib, Seaborn
Modeling TensorFlow, Keras
Environment Jupyter Notebook

🧠 Model Architectures

This project features two deep learning models built with TensorFlow/Keras: one for regression and one for classification. Both models share a clean, interpretable architecture and are optimized for educational data.

🔹 Regression Model — Predicting Exam Scores

Layer Configuration
Input Layer Receives preprocessed feature vector
Hidden Layer 1 Dense(256), ReLU activation
Dropout Layer Dropout(0.3)
Hidden Layer 2 Dense(128), ReLU activation
Hidden Layer 3 Dense(64), ReLU activation
Output Layer Dense(1)
  • Loss Function: Mean Squared Error (MSE)
  • Optimizer: Adam
  • Evaluation Metric: Mean Absolute Error (MAE)

This model predicts continuous exam scores based on behavioral and academic features.

🔹 Classification Model — Performance Tier Prediction

Layer Configuration
Input Layer Receives preprocessed feature vector
Hidden Layer 1 Dense(128), ReLU activation
Hidden Layer 2 Dense(64), ReLU activation
Output Layer Dense(3), Softmax activation
  • Loss Function: Categorical Crossentropy
  • Optimizer: Adam
  • Evaluation Metric: Accuracy

This model classifies students into three performance tiers: Low, Medium, and High

📈 Results & Evaluation

Model performance was evaluated using key metrics and visual diagnostics from the notebook.

📉 Regression Results

  • MSE: ~4.48
  • MAE: ~0.89
  • R² Score: ~0.69

Evaluation Visuals:

  • Predicted vs. Actual grade scatter plot

Actual vs Predicted

  • Training vs Validation Loss Curve

Training vs Validation Loss

  • Residual Distribution Plot

Residuals Distribution

📊 Classification Results

  • Training Accuracy: 100%
  • Validation Accuracy: ~98%

Evaluation Visuals:

  • Training vs Validation Accuracy

Training vs Validation Accuracy

The results indicate that both models generalize well, with stable learning curves and limited overfitting due to dropout and early stopping.

🚀 Installation

  1. Clone the repository
git clone https://github.com/ArianJr/student-performance-deep-learning.git
cd student-performance-deep-learning
  1. Install dependencies
pip install -r requirements.txt

🖥️ Usage

Open the Jupyter notebook to explore the analysis and models:

jupyter notebook student_performance_dl_analysis.ipynb

Follow the notebook to:

  • 📊 Explore data distributions
  • 🧠 Build and train models
  • 📈 Evaluate predictive performance

📊 Visualizations

Visual insights play a key role in understanding student performance:

  • Correlation Heatmap: Reveals relationships between features
  • Class Distribution: Shows balance of target labels
  • Model Metrics: Visualizes loss and accuracy trends

🔮 Future Improvements

  • 🧩 Feature Engineering: Explore polynomial or interaction features
  • 🧠 Hyperparameter Tuning: Use KerasTuner or Optuna for optimal architectures
  • 🧾 Cross-Validation: Add k-fold validation to reduce variance
  • 📊 Explainability: Incorporate SHAP or LIME for feature importance visualization
  • ⚙️ Deployment: Wrap models in a Flask API or Streamlit dashboard for real-time prediction
  • 📈 Data Expansion: Include demographic and attendance trends for improved accuracy

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

🙏 Acknowledgments

  • Dataset Source: Kaggle - Student Performance
  • Libraries: TensorFlow, Keras, Scikit-learn, Pandas, NumPy, Matplotlib, Seaborn
  • Inspired by educational data mining techniques and performance analytics research

👤 Author

Arian Jr
📧 Contact Me • 🌐 GitHub Profile

Made with ❤️ by ArianJr

⭐ If you found this project useful, please consider giving it a star! It helps others discover it and supports my work.

GitHub stars GitHub forks