Proj3: Air Quality

Overview

Proj3 analyzes air quality sensor data from Nairobi using time series techniques. The project progresses from data wrangling with MongoDB-style queries to autoregressive modeling.

Data Source

JSON sensor data with timestamps and PM2.5 readings.

Techniques

AR, ARMA, Linear Regression on lagged features

Evaluation

MAE, walk-forward validation, residual analysis

Lesson Workflows

Lesson	Topic	Technique	Query
3.1	Data Wrangling with MongoDB	Query, aggregate, reshape	`"run lesson 3.1"`
3.2	Linear Regression on Time Series	Lagged features + LinearRegression	`"run lesson 3.2"`
3.3	Autoregressive Models	AR(p) model fitting	`"run lesson 3.3"`
3.4	ARMA and Hyperparameter Tuning	ARMA(p,q) with grid search	`"run lesson 3.4"`

Data Structure

Proj3/data/ ├── nairobi.json # Raw sensor data └── cache/ ├── site_29_P2_wrangled.parquet ├── site_29_P2_wrangled_lagged.parquet ├── linear_regression_model_predictions.parquet └── autoreg_p1_model_predictions.parquet

Available Tools

Tool	Description
`load_proj3_data`	Load Nairobi air quality dataset
`analyze_air_quality`	Time series exploratory analysis
`forecast_pm25`	Generate PM2.5 forecasts
`fit_ar_model`	Fit autoregressive model
`fit_arma_model`	Fit ARMA model with tuning
`create_acf_plot`	Autocorrelation visualization

Quick Start

# Load and analyze "load nairobi air quality data" "analyze air quality time series" # Run specific lesson "run lesson 3.3 on AR models" # Forecast "forecast PM2.5 for next 24 hours" # Cross-project "apply GARCH from project 8 to air quality data"