Outline

1. Notation and basics for time series 

2. Basics of regression and forecasting 

   1. Fundamentals

   2. Autoregressive models

      1. 1. ARIMA

         2. SARIMA

3. Neural-network-based approaches for forecasting:  

   1. MLP

   2. CNN

   3. LSTM

   4. RNN

Resources

General course instructions and guidelines

Full material

Lectures

• Lecture 0: Motivation and course presentation

  • Slides

• Lecture 1: Basics of times series 

  • Slides

  • Lecture notes

• • Resources for the lecture:

    • Linear regression (Python Data Science Handbook  Notebook (HTML))

    • Time Series Forecasting with Python – Cheat Sheet 

    • Quick way to find p d and q values for arima

    • Guide to ac and pac plots in time series  

    • How to interpret acf and pacf plots PDF

    • MA, ARMA and ARIMA models in time series forecasting 

    • White noise and variance: Link 1  Link 2 

    • Time series example: ECG signal

• Lecture 2: Introduction to deep learning for time series forecasting

  • Slides

  • Lecture notes

  • Notebook

  • Other resources for the lecture:

    • Blog: How to Develop Multilayer Perceptron Models for Time Series Forecasting

    • The Multilayer Perceptron - Theory and Implementation of the Backpropagation Algorithm 

    • Framing time series problem as supervised-learning 

• Lecture 3: Introduction to Recurrent Neural Networks (RNNs)

  • Slides

  • Lecture Notes

  • Notebook

• Lecture 4: Long Short Term Memory (LSTM)

  • Slides

  • Lecture Notes

  • Notebook

• Lecture 5: Convolutional Neural Networks (CNN) for Time Series

  • Slides

  • Lecture Notes

  • Notebook

• Lecture 6: Summary and final remarks

  • Slides

• (Extra) Lecture: Summary

  • Slides

  • Notebook

  • Datasets:

    • Air-quality-London-time-of-day

    • Air-quality-London-monthly-averages

• Another example: Sport Car Price Prediction using Decision Tree Regressor, AdaBoost Regressor, and Random Forest Regressor

  • Google Colab Link 

  • Dataset

  • YouTube Channel : https://lnkd.in/gwmbVGUr

Resources:

• ARIMA model - Python complete guide 

• Orbit ML

• Comparison of benchmark forecasting techniques  (Python - HTML)

• Skforecast: time series forecasting with Python, Machine Learning and Scikit-learn (link)

• 𝐏𝐲𝐂𝐚𝐫𝐞𝐭 - 𝐍𝐞𝐰 𝐓𝐢𝐦𝐞 𝐒𝐞𝐫𝐢𝐞𝐬 𝐌𝐨𝐝𝐮𝐥𝐞 (𝐁𝐞𝐭𝐚)

  • 𝐃𝐨𝐜𝐮𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧

  • 𝐒𝐨𝐮𝐫𝐜𝐞 𝐜𝐨𝐝𝐞

  • 𝐑𝐨𝐚𝐝𝐦𝐚𝐩

  • 𝐄𝐱𝐚𝐦𝐩𝐥𝐞𝐬:

    • 𝐉𝐮𝐩𝐲𝐭𝐞𝐫 𝐧𝐨𝐭𝐞𝐛𝐨𝐨𝐤

    • 𝐆𝐨𝐨𝐠𝐥𝐞 𝐂𝐨𝐥𝐚𝐛 𝐧𝐨𝐭𝐞𝐛𝐨𝐨𝐤

• DeepAR

• Performance evaluation toolbox for time series models (Git)

• How to Use XGBoost for Time Series Forecasting (Link) 

• tsflex (Git): flexible time-series operations - Python toolkit for processing & feature extraction, making few assumptions about input data.

• Course:Time Series Analytics and Forecasting by Lorenzo Nespoli (Link).

• Initializing neural networks (Link).

• Amazon - Chronos:

  • Paper

  • Repo

• Example: Algorithms of Air Quality Estimation: A Comparative Study of Stochastic and Heuristic Predictive Models

• • Full paper 

  • Video 

  • Python notebook

Mid-term exam

• Project ideas:

  • Datasets

Assignments

• Part I: Basics of time series

  • Questionnaire

  • Notebook 

  • Datasets:

    • Air-quality-London-time-of-day

    • Air-quality-London-monthly-averages

• Part II: Long Short Term Memory (LSTM)

  • Questionnaire

  • Notebook

  • Dataset: Gold Price

• Part III: Introduction to Convolutional Neural Networks (CNNs):

  • Questionnaire

  • Notebook

  • Dataset: Climate