🎓 Student Success Prediction Dashboard

A beautiful Streamlit application that predicts student academic outcomes (Graduate, Dropout, or Enrolled) using a Random Forest model trained on 36 features including academic performance, demographics, and economic indicators.

🚀 Quick Start

1. Train and Save the Model

First, run the training script to create the pickle files:

python train_and_save_model.py

This will create:

• random_forest_model.pkl - The trained Random Forest model
• preprocessing_info.pkl - Preprocessing pipeline information
• feature_names.pkl - Feature names for consistency

2. Install Dependencies

pip install -r requirements.txt

3. Launch the Dashboard

streamlit run app.py

The dashboard will open in your browser at http://localhost:8501

📊 Model Performance

• Algorithm: Random Forest Classifier (n_estimators=100)
• Training Accuracy: 100%
• Test Accuracy: 93.84%
• Features: 36 student characteristics
• Classes: Graduate (50%), Dropout (32%), Enrolled (17%)

🎯 Features

Complete Input Form

• Personal Information: Marital status, gender, age, nationality
• Academic Background: Application details, course, previous qualifications
• Family Background: Parents' education and occupation
• Financial & Social: Scholarships, debts, special needs status
• Academic Performance: 1st and 2nd semester curricular units and grades
• Economic Indicators: Unemployment rate, inflation rate, GDP

Beautiful Interface

• Responsive Design: Clean, professional layout with custom CSS
• Interactive Widgets: User-friendly dropdowns with human-readable labels
• Real-time Predictions: Instant results with confidence scores
• Visual Analytics: Interactive charts showing prediction probabilities
• Model Insights: Performance metrics and key factors display

Accurate Preprocessing

• Exact Pipeline Match: Preprocessing mirrors the training notebook exactly
• One-hot Encoding: Categorical variables properly encoded
• Feature Consistency: Ensures all 105+ features match training data
• Input Validation: Handles missing features and maintains column order

🔧 Technical Details

Data Pipeline

1. Data Loading: Reads from ./data/data.csv
2. Column Cleaning: Removes apostrophes, fixes typos (Nacionality → Nationality)
3. Target Encoding: Maps Dropout=0, Enrolled=1, Graduate=2
4. One-hot Encoding: Converts categorical features to binary columns
5. Feature Alignment: Ensures prediction input matches training features

Model Architecture

• Random Forest with 100 estimators
• No scaling required (tree-based model)
• Train/Test Split: 70/30 with random_state=42
• Cross-validation: 5-fold validation during training

User Experience Enhancements

• Human-readable Labels: Maps numeric codes to descriptive text
• Smart Defaults: Pre-filled with typical student values
• Interactive Sliders: Easy input for numeric values
• Color-coded Results: Green for Graduate, Red for Dropout, Yellow for Enrolled
• Confidence Display: Shows probability scores for all three outcomes

📁 Project Structure

predictive_analysis/
├── data/
│   └── data.csv                    # Original dataset (4,424 students)
├── src/                            # Source code modules
│   ├── config/                     # Configuration files
│   │   ├── __init__.py
│   │   ├── theme.py               # Rose Pine theme and constants
│   │   └── mappings.py            # Feature mappings and presets
│   ├── utils/                     # Utility functions
│   │   ├── __init__.py
│   │   └── model_utils.py         # Model loading and preprocessing
│   ├── components/                # UI components
│   │   ├── __init__.py
│   │   ├── form_components.py     # Input form sections
│   │   └── prediction_components.py # Results and charts
│   └── __init__.py
├── predictive_analysis.ipynb      # Original analysis notebook
├── train_and_save_model.py        # Model training and pickle export
├── streamlit_app.py               # Legacy monolithic app (deprecated)
├── app.py                         # New modular main application
├── requirements.txt               # Python dependencies
├── random_forest_model.pkl        # Trained model (generated)
├── preprocessing_info.pkl         # Preprocessing pipeline (generated)
├── feature_names.pkl              # Feature list (generated)
└── README.md                      # This file

🎨 Dashboard Screenshots

The dashboard features:

• Sidebar Input Form: All 36 features organized by category
• Main Prediction Area: Large, colorful prediction display
• Probability Charts: Interactive bar charts and radar plots
• Model Information: Performance metrics and key insights
• Responsive Design: Works on desktop and mobile devices

🔍 Key Features Explained

Categorical Mappings

• Marital Status: Single, Married, Divorced, etc.
• Application Mode: Various admission pathways
• Course: Animation, Engineering, Management, etc.
• Qualifications: Education levels from basic to doctorate
• Occupations: Professional categories for parents

Academic Performance Tracking

• Curricular Units: Credits, enrollments, evaluations, approvals
• Grades: Average performance in 1st and 2nd semesters
• Without Evaluations: Units not evaluated

Economic Context

• Unemployment Rate: Regional economic indicator
• Inflation Rate: Economic stability measure
• GDP: Overall economic performance

🚀 Deployment Options

Local Development

streamlit run streamlit_app.py

Streamlit Cloud

1. Push code to GitHub
2. Connect repository to Streamlit Cloud
3. Deploy with automatic requirements detection

Docker

FROM python:3.9-slim
COPY . /app
WORKDIR /app
RUN pip install -r requirements.txt
EXPOSE 8501
CMD ["streamlit", "run", "streamlit_app.py"]

🛠️ Customization

Adding New Features

1. Update train_and_save_model.py with new preprocessing
2. Modify streamlit_app.py input form
3. Ensure feature alignment in preprocess_input()

Styling Changes

• Modify CSS in the st.markdown() sections
• Update color schemes in color_discrete_map
• Adjust layout in column configurations

Model Improvements

• Tune hyperparameters in RandomForestClassifier()
• Add feature selection or engineering
• Implement cross-validation strategies

📊 Model Validation

The model achieves excellent performance:

• ROC AUC: 0.98 (nearly perfect discrimination)
• Precision: Graduate 95%, Dropout 92%
• Recall: Graduate 95%, Dropout 93%
• F1-Score: Balanced across all classes

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Submit a pull request

📄 License

This project is open source and available under the MIT License.

🙏 Acknowledgments

• Dataset from UCI Machine Learning Repository
• Built with Streamlit, scikit-learn, and Plotly
• Inspired by educational data mining research