An end-to-end Machine Learning classification project designed to analyze student sleep patterns, lifestyle factors, and predict mental health outcomes. Multiple algorithms were evaluated alongside Grid/Randomized Search hyperparameter tuning to achieve an optimized predictive accuracy of ~82.34%.
Student-Sleep-Mental-Health-Prediction/
├── student_sleep_mental_health_2026.csv # Dataset containing student sleep & mental health metrics
├── model.ipynb # Jupyter Notebook with EDA, Hyperparameter Tuning & Evaluation
├── model.pkl # Best performing trained classification model
├── scaler.pkl # Feature scaling object for normalization
├── columns.pkl # Saved column structure to ensure input alignment
└── README.md # Project documentation
🛠️ Hyperparameter Tuning Details (XGBoost)
Systematic hyperparameter search was conducted across 5-fold cross-validation (split0 to split4) to optimize regularizations and sampling parameters (colsample_bytree, subsample, reg_lambda, reg_alpha):
Best Mean Test Score: 0.823383 (~82.34%)
Standard Deviation (std_test_score): 0.006858(high model stability across folds)
⚡ Key Highlights & Methodology
Interactive Streamlit Web App: Built a user-friendly UI for real-time mental health score prediction based on user inputs.
Data Preprocessing & Scaling: Applied StandardScaler (scaler.pkl) to ensure numerical feature compatibility.
Feature Schema Alignment: Serialized columns.pkl to safeguard exact feature ordering during real-time inference.
Systematic Cross-Validation: Utilized 5-fold cross-validation to prevent overfitting on student behavioral metrics.
Deployment Ready: Exported all essential pipeline artifacts (model.pkl, scaler.pkl, columns.pkl).
End-to-end Machine Learning project predicting student mental health from sleep & lifestyle metrics. Features EDA, 5-fold CV hyperparameter tuning (XGBoost/Logistic Regression ~83% accuracy), and an interactive Streamlit UI web app.
Student Sleep & Mental Health Prediction (2026)
An end-to-end Machine Learning classification project designed to analyze student sleep patterns, lifestyle factors, and predict mental health outcomes. Multiple algorithms were evaluated alongside Grid/Randomized Search hyperparameter tuning to achieve an optimized predictive accuracy of ~82.34%.
Student-Sleep-Mental-Health-Prediction/
├── student_sleep_mental_health_2026.csv # Dataset containing student sleep & mental health metrics
├── model.ipynb # Jupyter Notebook with EDA, Hyperparameter Tuning & Evaluation
├── model.pkl # Best performing trained classification model
├── scaler.pkl # Feature scaling object for normalization
├── columns.pkl # Saved column structure to ensure input alignment
└── README.md # Project documentation
🛠️ Hyperparameter Tuning Details (XGBoost)
Systematic hyperparameter search was conducted across 5-fold cross-validation (split0 to split4) to optimize regularizations and sampling parameters (colsample_bytree, subsample, reg_lambda, reg_alpha):
Best Mean Test Score: 0.823383 (~82.34%)
Standard Deviation (std_test_score): 0.006858(high model stability across folds)
⚡ Key Highlights & Methodology
Interactive Streamlit Web App: Built a user-friendly UI for real-time mental health score prediction based on user inputs.
Data Preprocessing & Scaling: Applied StandardScaler (scaler.pkl) to ensure numerical feature compatibility.
Feature Schema Alignment: Serialized columns.pkl to safeguard exact feature ordering during real-time inference.
Systematic Cross-Validation: Utilized 5-fold cross-validation to prevent overfitting on student behavioral metrics.
Deployment Ready: Exported all essential pipeline artifacts (model.pkl, scaler.pkl, columns.pkl).
End-to-end Machine Learning project predicting student mental health from sleep & lifestyle metrics. Features EDA, 5-fold CV hyperparameter tuning (XGBoost/Logistic Regression ~83% accuracy), and an interactive Streamlit UI web app.
