INN358 Machine Learning with Discrete Event Stream Data
Norsk emneinformasjon
Search for other courses here
Select other year
Showing course contents for the educational year 2021 - 2022 .
Course responsible: Joachim Scholderer
Teachers: Mike Riess
ECTS credits: 5
Faculty: School of Economics and Business
Teaching language: EN
(NO=norsk, EN=Engelsk)
(NO=norsk, EN=Engelsk)
Limits of class size:
70
Teaching exam periods:
This course will be offered first time autumn 2022. This course starts in the Autumn parallel. This course has teaching/evaluation in the Autumn parallel.
Course frequency: Annually
First time: Study year 2021-2022
Preferential right:
- Study specialisations where this course is mandatory have priority: digital business transformation (M-EI), business analytics (M-TDV) and business analytics (M-ØA)
- Students from data science (M-DV and M-TDV), entrepreneurship and innovation (M-EI), industrial economics and technology management (M-IØ) and business administration (M-ØA) who have passed all mandatory and recommended prerequisites for the course
Course contents:
Most standard techniques in machine learning are designed for cross-sectional data. Discrete event stream data (e,.g., event log data from ERP, CRM and SCM systems) also have a time dimension, which leads to complex dependency structures in the data. In this course, participants will learn advanced modelling and prediction techniques that can handle such data.
- Dependency structures in discrete event stream data
- Pre-processing and feature engineering
- Use of standard machine learning techniques with aggregated event stream data
- Process mining, association rules and link analysis
- Time-to-event models
- Hidden Markov models
- Dynamic Bayesian networks
- Model assessment and selection
- Implementation in Python, SAS and R
Practical work with real cases is an important part of the course. Participants will work in groups on a semester-long case project.
Learning outcome:
Knowledge
- Understand the dependency structures in discrete event stream data
- Understand the theoretical foundations of modelling and prediction techniques that can handle discrete event stream data
Skills
- Be able to perform descriptive analyses of discrete event stream data
- Be able to develop prediction models based on discrete event stream data
- Be able to evaluate models and choose between competing models
General competence
- Be able to work in cross-functional project structures
- Be able to contribute constructively in process automation projects
Learning activities:
Lectures, exercises with data and software, case workshops under supervision, independent group work related to the project assignment
Teaching support:
Canvas, Microsoft Teams
Syllabus:
Tutz, G., & Schmid, M. (2016). Modeling discrete time-to-event data. Cham: Springer.Selected journal articles and book chapters
Prerequisites:
Recommended prerequisites:
- INN265 Analysis of business processes
Assessment:
Portfolio assessment , consisting of an individual midway home exam (weight. 50%) and a project assignment conducted in groups of four participants (weight: 50%). No re-sit examination will be arranged in this course.
Nominal workload:
125 hours
Reduction of credits:
INN355 (5 ECTS)
Type of course:
- Lectures and case workshops: 20 hours
- Exercises with data and software: 5 hours
- Project assignment: 50 hours
- Self-study/syllabus literature: 50 hours
Examiner:
External examiner will control the quality of the syllabus, questions for the examination, and principles for the assessment of the examination answers.
Examination details: :