## Bike Pollution Regression & Classification Pipeline
This Airflow DAG performs an end-to-end ETL, feature engineering, and modeling pipeline for predicting bike rentals based on weather, air pollution, and historical bike rental data. It includes both regression and classification models.

The pipeline extracts historical bike rental data and daily air pollution measurements, merges them, computes relevant features, and trains multiple ML models to predict:
    Next-hour bike rentals (regression)
    High-demand hours (classification)

## Regression models:
    Linear Regression
    Random Forest Regressor

## Classification models:
    Logistic Regression
    Support Vector Machine (SVM)

The DAG automatically selects the best model for each task based on evaluation metrics (MAE for regression, F1-score for classification) and generates a final summary JSON.

## Features
    Automatic extraction of bike rental CSV and EPA AQS daily pollution data
    Feature engineering including lag features, rolling averages, peak hours, and weekend indicators
    Train/test split with time-based cutoff
    ML preprocessing pipeline: scaling numeric features, one-hot encoding categorical features
    Multiple regression and classification models with automated model selection
    Final summary JSON containing best models and classification threshold