🎓 Student Performance Prediction using Machine Learning

This project focuses on building, analyzing, and evaluating a Machine Learning model to predict student exam performance based on academic, behavioral, and socio-economic factors.

Unlike UI-centric applications, this project emphasizes the complete ML lifecycle, including data preprocessing, feature engineering, model comparison, diagnostics, and ethical considerations.

---

📌 Problem Statement

To develop a predictive model that estimates a student's final exam score using structured educational and behavioral data.

The objective is not only to generate predictions but also to understand:

• Which features influence performance
• How model assumptions affect results
• The limitations and ethical implications of predictive modeling in education

---

🔁 Machine Learning Lifecycle

1️⃣ Data Cleaning

• Checked for missing values
• Verified data types
• Removed inconsistent or duplicate records
• Ensured unit consistency (e.g., study hours per week)
• Validated numerical ranges (attendance %, scores, etc.)

---

2️⃣ Encoding

Since the dataset contains categorical variables:

• Ordinal Encoding used for:

  • Parental Involvement
  • Motivation Level
  • Access to Resources
  • Family Income
  • Teacher Quality
  • Parental Education Level

• Binary Encoding used for:

  • Internet Access
  • Extracurricular Activities
  • Learning Disabilities
  • Gender

• One-Hot Encoding used for:

  • School Type
  • Peer Influence

Encoding was carefully designed to avoid introducing artificial order in nominal variables.

---

3️⃣ Outlier Analysis

• Examined numerical features using statistical summaries
• Identified extreme values in:
  • Study Hours
  • Previous Scores
  • Attendance
• Verified whether outliers were:
  • Data entry errors (removed)
  • Genuine rare cases (retained)

Outlier handling ensured model stability without distorting real-world variance.

---

4️⃣ Model Training & Comparison

Multiple models were evaluated:

• Linear Regression
• Random Forest Regressor
• Gradient Boosting Regressor

Evaluation metrics used:

• R² Score
• Mean Absolute Error (MAE)
• Mean Squared Error (MSE)

Model comparison allowed selecting the most interpretable and stable model for deployment.

Final selected model: Linear Regression (for interpretability and consistency).

---

5️⃣ Model Diagnostics

To ensure reliability:

• Checked residual distribution
• Analyzed prediction spread across low and high score ranges
• Verified absence of extreme bias toward mean
• Evaluated generalization performance on test data

Observed behavior:

• Slight regression toward the mean (expected in linear models)
• Stable prediction performance across moderate score ranges

---

6️⃣ Ethical Reasoning

Educational prediction systems must be handled responsibly.

This project acknowledges:

• Predictions are probabilistic, not deterministic
• Socio-economic factors must not reinforce bias
• Model outputs should not label or restrict student potential
• The system is for analytical and educational research purposes only

Ethical deployment requires:

• Transparency
• Bias awareness
• Responsible interpretation

---

📊 Features Used

Academic Factors

• Hours Studied
• Attendance
• Previous Scores
• Sleep Hours
• Tutoring Sessions
• Physical Activity

Socio-economic & Behavioral Factors

• Parental Involvement
• Access to Resources
• Family Income
• Teacher Quality
• Peer Influence
• Internet Access
• Extracurricular Activities
• Learning Disabilities
• Gender
• Distance from Home

---

🛠️ Tech Stack

Language: Python

Libraries:

• scikit-learn
• pandas
• numpy
• streamlit
• matplotlib

Model: Linear Regression
Deployment Platform: Streamlit Community Cloud

---

📂 Project Structure

LEVEL1_StudentPerformancePredictor/
│
├── app.py  #Streamlit UI for checking prediction
|
├── student_score_model.pkl  #Linear Regression Model for predicting the final score
│
├── requirements.txt
├── README.md

---

🚀 Deployment

The application is deployed using Streamlit Community Cloud, allowing seamless hosting of the Streamlit application directly from GitHub.

Deployment Process:

1. Push the complete project to a public GitHub repository.
2. Add requirements.txt for dependency management.
3. Connect the repository to Streamlit Community Cloud.
4. Deploy app.py as the entry point.

Live Application:
🔗 App URL: https://level1studentperformancepredictor-m2wfrbrej9jhxmuutaufe3.streamlit.app/

---

🚀 Future Enhancements

• Inclusion of Social Media Usage as a behavioral factor
• Hyperparameter tuning for ensemble models
• Feature importance visualization
• Bias detection analysis
• Cross-validation-based stability testing

---

👩‍💻 Author

Apeksha Machine Learning and Python enthusiast

---

📜 License

This project is intended for educational and academic purposes.