Course Abstract

Despite being not the youngest branch of data analysis, time series forecasting still poses a great challenge to both researchers and practitioners. As Niels Bohr said years ago “Prediction is very difficult, especially when it’s about the future”. Fortunately, plurality of approaches have been proposed to address this commonly appearing challenge. This course introduces the users to the most prominent and widely used solutions, explaining their advantages and disadvantages together with tips and recommendations on the suited-for-purpose model usage. To facilitate rapid transition of time series theory into actual business applications that students may encounter and profit from in real life, the course is equipped with hands-on code run-throughs provided in python. At the end of the course, the prediction will for sure be less difficult.

DIFFICULTY LEVEL: INTERMEDIATE

Key Skills and Tools

  • Python

Learning Objectives

  • Understand the essential theory of both basic and advanced time series models

  • Build production-ready time series forecasts with python libraries

  • Interpret the output of time series models to transform them into business insights

Instructor Bio:

Leonardo De Marchi

SENIOR DATA SCIENTIST | LECTURER | InPost | Wrocław University of Economics and Business

Marta Markiewicz

Currently Senior (Big) Data Scientist at InPost and Lecturer at Wroclaw University of Economics and Business, previously Head of Data Science at Objectivity, with a background in Mathematical Statistics. For almost 10 years, she has been discovering the potential of data in various business domains, from medical data, through retail, HR, finance, aviation, real estate, logistics, … She deeply believes in the power of data in every area of life. Articles’ writer, conference speaker and privately – passionate dancer and hand-made jewellery creator.

Course Outline

Module 1: Time Series Introduction 

  • Course agenda
  • Real-life examples
  • Time series definition
  • Example in python


Module 2: Benchmark methods

  • Distribution focused
  • Naive approaches
  • Expert forecasts
  • Example in python


Module 3: Performance evaluation techniques 

  • Purpose of performance evaluation
  • Useful metrics
  • Evaluation techniques
  • Example in python


Module 4: Exponential smoothing 

  • Single exponential smoothing
  • Holt’s linear trend model
  • Holt-Winters exponential smoothing
  • Example in python


Module 5: (S)AR(I)MA(X)

  • AR
  • MA
  • ARIMA
  • SARIMA
  • SARIMAX
  • Example in python


Module 6: (Linear) regression 

  • Linear regression
  • Support Vector Machines
  • Trees: Decision Trees, Random Forest and Boosting
  • Example in python


Module 7: Prophet 

  • Prophet
  • Example in python


Module 8: Neural Networks 

  • Artificial Neural Networks
  • Recurrent Neural Networks
  • LSTM
  • TCN
  • Example in python


Module 9: Tricks that improve model performance 

  • Outliers
  • Fourier series
  • Hierarchical reconciliation
  • Time reconciliation
  • Power transform
  • Example in python


Module 10: Course wrap-up

  • Summary of covered methods and libraries


Background knowledge

  • Some experience with machine learning would make this workshop easier to follow, but is by no means necessary

  • All code demos during the training will be in Python, so experience with it or another similar programming language would be helpful

Use-cases this course could be useful for

  • Finance: Forecasting revenue

    Time series forecasting is used by various companies to predict the revenue - either globally or for certain branches/regions. Accurate revenue forecasting supports tough decision-making and makes planning easier.

  • Retail : Forecasting sales

    Based on historical sales and additional phenomena like promotions or the current market situation, it’s possible to accurately predict short or long-term sales, estimate the impact of promotion, and check what-ifs scenarios.

  • HR: Forecasting staff turnover

    Time series supports HR with the ability to forecast and as a bonus - to better understand staff turnover. It helps the company to prepare for, or even prevent, employees leaving. As a consequence, the cost associated with the necessity of recruiting new staff as well as onboarding is decreasing.

  • Software developers, data scientists, analysts, statisticians, and other data-related professionals are the core target audience for this training. This training is for anyone who would like to create the best possible forecasts for real life time series, regardless of business domain.