Python for Time Series Notes

Study notes and interactive Jupyter notebooks for Time Series Analysis and Forecasting — from fundamentals through ARIMA, Exponential Smoothing, Prophet, Deep Learning, and modern forecasting frameworks. All content is delivered through executable Jupyter notebooks designed for self-paced learning.

Part of the Python Learning Series: Data Science · Time Series · Deep Learning (planned) · NLP (planned)

Requirements

• Python 3.12+
• Poetry

Quick Start

# 1. Install dependencies
make install

# 2. Open notebooks for a chapter
make jupyter CH=01

Project Structure

python-for-time-series-notes/
├── src/
│   ├── data/                # Shared datasets (CSV)
│   ├── chapter_01/          # Getting Started with Time Series
│   ├── chapter_02/          # Time Series with Pandas
│   ├── chapter_03/          # Time Series Visualization
│   ├── chapter_04/          # Time Series Decomposition
│   ├── chapter_05/          # Time Series Features & Statistics
│   ├── chapter_06/          # The Forecaster's Toolbox
│   ├── chapter_07/          # Exponential Smoothing
│   ├── chapter_08/          # ARIMA Models
│   ├── chapter_09/          # Dynamic Regression (SARIMAX)
│   ├── chapter_10/          # Multivariate Models (VAR/VARMA)
│   ├── chapter_11/          # Prophet
│   ├── chapter_12/          # Deep Learning for Time Series
│   ├── chapter_13/          # Modern Forecasting Frameworks
│   └── chapter_14/          # Capstone Project
├── tests/                   # Pytest test suite
├── docs/                    # Reference books (gitignored)
├── Makefile                 # Automation commands
└── pyproject.toml           # Poetry configuration

Chapters

Part I: Time Series Fundamentals

Chapter	Notebooks	Topics
01 - Getting Started	3	What is a time series, types of data, Python datetime, components (trend, seasonality, noise), the forecasting workflow
02 - Time Series with Pandas	5	DatetimeIndex, frequency and periods, resampling, time shifting, rolling and expanding windows, missing data (ffill, bfill, interpolate)
03 - Visualization	3	Time plots, seasonal plots, subseries plots, lag plots, ACF plots, matplotlib date formatting
04 - Decomposition	4	Transformations (Box-Cox, log), moving averages, classical decomposition, STL decomposition, additive vs multiplicative
05 - Features & Statistics	4	Summary statistics, ACF and PACF theory, stationarity (ADF, KPSS tests), differencing, Granger causality, white noise

Part II: Statistical Forecasting

Chapter	Notebooks	Topics
06 - The Forecaster's Toolbox	4	Simple methods (naive, drift, mean, seasonal naive), residual diagnostics, prediction intervals, evaluation metrics (MAE, RMSE, MAPE), time series cross-validation
07 - Exponential Smoothing	4	SES, Holt's linear trend, Holt-Winters, ETS state-space taxonomy, damped trends, model selection (AIC/BIC)
08 - ARIMA Models	5	Stationarity and differencing, backshift notation, AR, MA, ARIMA, seasonal ARIMA, auto_arima, ARIMA vs ETS
09 - Dynamic Regression	3	Regression with ARIMA errors, exogenous variables (SARIMAX), harmonic regression, lagged predictors
10 - Multivariate Models	3	VAR, VARMA, Granger causality for variable selection, impulse response functions, cointegration

Part III: Modern Forecasting

Chapter	Notebooks	Topics
11 - Prophet	3	Model internals (piecewise trend, Fourier seasonality), changepoints, holidays, cross-validation, evaluation
12 - Deep Learning	4	RNN/LSTM architecture and gates, GRU, sequence-to-sequence, Temporal Fusion Transformer, N-BEATS overview
13 - Modern Frameworks	3	sktime unified API, Nixtla statsforecast/mlforecast, model comparison, forecast combinations
14 - Capstone Project	2	End-to-end forecasting: EDA, decomposition, multiple models, evaluation, model selection

Running Notebooks

# Open a specific chapter in Jupyter Lab
make jupyter CH=08

# List all chapters
make list-chapters

# List notebooks in a chapter
make list-notebooks CH=07

Code Quality

make lint           # Ruff linter
make format         # Auto-format
make type-check     # mypy
make check          # All checks

Development

See CONTRIBUTING.md for workflow, commit conventions, and code style guidelines.

References

• Jose Portilla's Python for Time Series Data Analysis (Udemy) — practical exercises and datasets
• Forecasting: Principles and Practice, 3rd Edition — Rob J Hyndman & George Athanasopoulos — theoretical foundation
• statsmodels Documentation
• sktime Documentation
• Nixtla statsforecast Documentation
• Prophet Documentation
• pmdarima Documentation
• Keras 3 Documentation

Author

Leandro Mana

License

MIT