Longevity Prediction

Predict lifespan from lifestyle habits using four regression models with GridSearchCV hyperparameter tuning, plus EDA and diagnostic visualizations.

Overview

This project builds a regression pipeline to predict age at death from features like work hours, sleep, exercise, and occupation type. GridSearchCV tunes each model's hyperparameters via 3-fold cross-validation. All training runs on the backend; the frontend renders metrics and plots without a page reload.

Models

Model	GridSearchCV parameters
Random Forest	n_estimators, max_depth, min_samples_split
Gradient Boosting	n_estimators, learning_rate, max_depth
Ridge Regression	alpha (0.1, 1.0, 10.0, 100.0)
Lasso Regression	alpha (0.01, 0.1, 1.0, 10.0)

Features

• EDA plots — correlation heatmap, stacked histogram by occupation, PairGrid dot plot
• Metrics — test/train R², MAE, MSE, RMSE, cross-validation R², best hyperparameters
• Diagnostic plots — prediction scatter, feature importance, residuals vs fitted, residuals distribution, learning curve
• Caching — preprocessed data and trained models are cached in memory so re-running is instant

Dataset

"Quality of Life" dataset with lifestyle and demographic features:

Feature	Type
work_hours	Numeric
rest_hours	Numeric
sleep_hours	Numeric
exercise_hours	Numeric
occupation_type	Categorical (one-hot encoded)
gender	Categorical (one-hot encoded)
age_at_death	Target

Train/test split: 80/20. Numeric features scaled with StandardScaler.

Tech Stack

• Backend: Flask, scikit-learn, pandas, numpy, matplotlib, seaborn
• Frontend: Chart.js 3.9.1, Vanilla JavaScript

Project Structure

work_life_regression/
├── __init__.py                       # Flask blueprint, training pipeline, GridSearchCV, API routes
├── templates/
│   └── work_life_regression/
│       └── index.html                # Model selector, metrics grid, EDA + diagnostic plot panels
└── static/
    └── script.js                     # Fetch calls, Chart.js feature importance bar chart

API

Method	Path	Description
GET	/work-life-regression/	Main page
GET	/work-life-regression/run?model=<key>&force=<bool>	Train a model, return metrics + plots
GET	/work-life-regression/plots	Generate 3 EDA plots as base64 PNG
GET	/work-life-regression/diagnostics?model=<key>	Generate learning curve for a model

Model keys: random_forest, gradient_boosting, ridge, lasso