INN355 Machine Learning for Business Process Optimisation
Check for course changes due to the coronavirus outbreak on Canvas and StudentWeb.
Showing course contents for the educational year starting in 2019 .
Course responsible: Joachim Scholderer
Teachers: Mike Riess
ECTS credits: 10
Faculty: School of Economics and Business
Teaching language: EN
Teaching exam periods:
This course starts in the Spring parallel. This course has teaching/evaluation in the Spring parallel.
Course frequency: Annual
First time: 2017H
Digitalisation and process automation are the hallmark of Industry 4.0. Machine learning plays a prominent role in this, providing the algorithmic and statistical "toolbox" for the automation and data-driven optimisation of business processes. The course has a two aims. The first aim is to introduce participants to the machine learning methodologies and data science tools that are central in business applications:
- Data visualisation and reporting (dimension reduction, plotting techniques, dashboard design),
- Predictive analytics (regularised least squares techniques, nearest neighbour, decision trees, ensemble methods such as bagging and boosting),
- Process analysis (time-to-event modelling, statistical process control, anomaly detection, process mining)
- Model assessment and selection (bias-variance trade-off, partitioning and cross-validation, decision theory, cost-sensitive learning).
The biggest risk in applied machine learning projects is a disconnect between the orderly world of data science and the messy reality of business processes. Hence, the second aim is to cultivate participants¿ ability to manage machine-learning projects in real business contexts:
- Requirements engineering (strategic and operational context, collaboration with project stakeholders, elicitation and specification of requirements, life cycle management),
- Data management (enterprise architecture, data protection and security, data management plans, data quality, feature engineering),
- Model deployment (implementation, performance management, error handling, updating).
Hands-on work on real and current business cases is a key part of the course. Participants will work in teams on two case challenges, organised in close cooperation with local and international businesses.
- Understand the capabilities of important machine learning techniques
- Know how to leverage these techniques to automate decisions and optimise business processes
- Know how to train, tune and test models for classification, prediction and anomaly detection
- Be able to choose the best model from among several competing models
- Know how to deploy machine learning models in typical business process flows
- Be able to use data visualisation tools such as Tableau and self-service analytics tools such as Power BI to develop dashboards and reporting systems
- Be able to use high-level platforms such as SAS Enterprise Miner and Celonis to build models for classification, prediction and anomaly detection
- Be able to manage process automation projects in a wide range of business contexts,
- Understand the need to align model deployment with data management and enterprise architecture
- Be able to work in cross-functional project structures
- Understand and manage ethical and regulatory issues in digital business contexts
Lectures, workshops and tutorials under supervision, flipped classroom activities, assignments and independent teamwork related to the case challenges.
Hastie, T., Tibshirani, R. & Friedman, J. (2008). The elements of statistical learning: Data mining, inference and prediction (2nd Ed.). Berlin: Springer. (E-book download and companion website: http://www-stat.stanford.edu/~tibs/ElemStatLearn/)
Van der Aalst, W. (2016). Process mining: Data science in action (2nd Ed.). Berlin: Springer.
In addition, the readings will include a selection of journal articles and case studies. The more tool-oriented parts of the course programme will be supported by online tutorials. Details will be announced on the Canvas page of the course at the beginning of the semester.
STAT100 Statistics or equivalent
MATH100 Introductory Mathematics
STAT200 Regression Analysis or ECN201 Econometrics
Two project assignments (weight: 50% each), conducted in groups of four participants, related to the case challenges. No re-sit examination will be arranged.
300 hours. This is a very work-intensive course.
3rd year (bachelor) or higher
Reduction of credits:
DAT200 Applied machine learning (credit reduction: 5 ECTS)
Type of course:
Four lecture hours per week (February-May). In addition, intensive case work in project teams.
The course is in English. Incoming students can contact student advisors at the School of Economics and Business (firstname.lastname@example.org) for admission to the course.
External examiner will control the quality of syllabus, questions for the final examination, and principles for the assessment of the examination answers.
Examination details: Continuous exam: A - E / F