AI Bootcamp: Time-series Forecasting

 

Time-series forecasting is a special edition of our intensive one-week AI Bootcamp Programme, designed for those looking to master the art of predicting future trends. From business and finance to supply chain management, energy, and healthcare, time-series forecasting plays a crucial role in strategic decision-making.

In this course, you’ll explore the fundamentals of time-series analysis, learning how to uncover patterns in data and build predictive models using both statistical methods and machine learning. But it’s not just theory — you’ll get hands-on experience with Python, working with real-world time series data to develop practical forecasting solutions.

 


Who it is for

This course is designed for business professionals looking to specialize in time-series data analysis and build their first statistical and machine learning models for forecasting. It is intended for those who already have experience in machine learning and a solid understanding of the data they work with. During the course, we provide a strong foundation in time-series analysis and modeling, equipping participants with the skills to confidently apply these techniques in their field.


Syllabus

  • time-series data processing, decomposition, analysis and visualization
  • statistical modeling (ETS, ARIMA, ...)
  • time-series feature engineering
  • machine learning modeling
  • multi-step and hierarchical forecasting
  • model selection and validation for time-series data
  • demonstration and hands-on case studies

What will you learn

  • how to analyze and plot time-series data in Python
  • how to efficiently process time-series data
  • to understand the basic time-series forecasting principles
  • how to tune and implement statistical and ML forecasting models
  • best practices in time-series data validation
    (cross-validation without data leakage, various metrics...)
  • apply forecasting techniques to real world data and scenarios

Prerequisites

  • understanding of ML principles and Python programming skills with focus on data processing

      OR

Pricing

  • 1.200 EUR per person (w/o VAT) includes:
    • one-week full-day course
    • course materials (code, datasets, presentations)
    • lunch
    • shareable certificate

Course schedule


Case studies

Demonstration

The demonstrative case study is based on a publicly available dataset and covers all key stages of a time-series forecasting project — from data preprocessing and analysis to model tuning, evaluation, and error analysis. Throughout the case study, baseline, statistical, and machine learning models are developed and validated, incorporating ML feature engineering, hyperparameter tuning, time-series cross-validation strategies, single- and multi-output approaches, model error analysis, and multiple validation metrics.

Hands-on

The hands-on case study takes place through a private Kaggle competition, focusing on one-week-ahead forecasting of electrical energy consumption for a simulated large-scale office building. Participants work with two years of historical consumption data, analyzing and preprocessing it to uncover detailed consumption patterns. Using this data, they independently develop hourly resolution forecasts for the upcoming week, with full support from course tutors throughout the process.

Contact

Address Laboratory for Renewable Energy Systems (LARES)
Faculty of Electrical Engineering and Computing
Unska 3
HR-10000 Zagreb
Croatia
Phone +385 1 6129 893
Email hrvoje.novak@fer.hr