PhD course: Business Analytics

Spring semester

Credits
5 ECTS

Teaching method
Classroom instruction.
14x2 lectures, 7 weeks, + 2 tutorials (2nd half of the semester).

Language
English

Examination
Individual oral examination, 30 min., with no aids and no preparation time. For this exam, the use of tools based on generative AI is not allowed. 

Assessment
Pass/No pass, internal co-examination.

Lecturers
Marcel Turkensteen, Chen Huang and Jenny Hongyan Li

Purpose

The purpose of the course is to provide PhD students with an understanding of how business analytics can support managerial decision-making through prediction, optimization, and the integration of the two. The course introduces central predictive and prescriptive approaches and highlights how these can be combined in data-driven decision processes under uncertainty. The course particularly aims to strengthen the students’ ability to connect analytical methods with relevant business applications and research questions.

Course description - DRAFT:

The course is split into 2 modules: 

Module one covers predictive modelling and causal machine learning for business analytics research, focusing on rigorous evaluation, interpretation, and uncertainty. Topics include risk minimization, regularization, tree-based and ensemble methods, model validation, and causal estimation.

Module two covers prescriptive analytics and optimisation, focusing on formulating optimisation problems from for business decision problems. Topics include linear and integer linear programming, solution procedures including both exact and heuristic methods, and validation of models and solutions. 

Module 3 covers the intricate interplay between prediction and prescription.  The module builds on the two previous modules by showing how optimisation can be used to translate predictions into prescriptions. The module emphasizes that accurate predictions do not necessarily lead to high-quality decisions, and pitfalls of the predict-then-optimise pipeline are explored.

Academic prerequisites
Students should have a basic understanding of programming in a statistical language (e.g., R or Python), as well as foundational knowledge in statistics, linear algebra, and optimization methods.

For PhD students at AU (ECON), this course has been pre-approved as an internal BSS PhD course.

Contents
Module 1:

  • Prediction as risk minimization: population and empirical risk, generalization error, and the bias–variance trade-off.
  • Regularization and high-dimensional models: ridge, lasso, elastic net, and hyperparameter tuning
  • Tree-based and ensemble methods: decision trees, random forests, and gradient boosting
  • Model evaluation and robustness: cross-validation design, data leakage, uncertainty assessment, and distribution shift
  • Causal machine learning: causal identification, treatment-effect estimation, doubly robust methods, and cross-fitting
  • From prediction and causal effects to business decisions: cost-sensitive learning, threshold optimization, uplift modelling, and policy evaluation

Module 2: 

  • Optimization as prescriptive analytics: optimizing the consequences of decisions in business analytics
  • Modelling for decision-making; model use and the purpose of the study (e.g., designing a solution method or generating managerial insights)
  • Core business optimization paradigms: (integer) linear programming, heuristics, analytical approaches
  • Modelling business decision problems: formulation of problems such as allocation, planning, scheduling, pricing, and supply chain problems
  • Experimentation and validation of decision models

Module 3: 

  • From prediction then optimization to integrated prediction and optimization
  • Big-data driven newsvendor models and solutions methods.
  • From quantile regression to distributional robust optimization.
  • Opportunities and challenges of integrated prediction and optimization

Each module includes programming applications.

Description of qualifications
Knowledge and understanding of:

  • of the foundations of statistical learning, including risk minimization, regularization, and modern machine learning methods
  • of principles for model evaluation, generalization, and robustness in empirical applications
  • the role of optimization as a prescriptive analytics approach in business analytics, and its relation to descriptive and predictive methods
  • central optimization paradigms used in business decision-making, including linear, integer, and network-based models
  • how prediction and optimization can be combined in data-driven decision-making 

Skills to:

  • to implement, tune, and evaluate predictive models using modern statistical software
  • to critically assess model performance, diagnose common pitfalls, and interpret results using explainability tools
  • formulate optimization models for business decision problems by defining decision variables, objective functions, and constraints
  • assess and select relevant optimization approaches for different types of business analytics problems
  • analyse and discuss how predictive models and optimization models can be integrated to support data-driven decision-making 

Competences to:

  • design and implement predictive modelling strategies for complex empirical problems
  • critically evaluate methodological choices and integrate predictive and causal approaches in applied settings
  • critically reflect on the use of optimization models as decision support tools in business analytics and assess their assumptions, strengths, and limitations
  • independently apply optimization to analyse complex business decision problems and justify relevant modelling choices
  • evaluate how predictive and prescriptive methods can be combined to support robust and informed managerial decision-making in uncertain environments

Literature
The lectures are based on selected chapters of textbooks and research papers. Potential textbooks and research articles include (but are not limited to):

Module 1:
James, G., Witten, D., Hastie, T., & Tibshirani, R. (2021). An Introduction to Statistical Learning: With Applications in R (2nd ed.). Springer. (relevant chapters: 2, 5-6, 8)

Cawley, G. C., & Talbot, N. L. C. (2010). On over-fitting in model selection and subsequent selection bias in performance evaluation. Journal of Machine Learning Research, 11, 2079–2107.

Athey, S., & Imbens, G. (2016). Recursive partitioning for heterogeneous causal effects. Proceedings of the National Academy of Sciences, 113(27), 7353–7360.

Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W., & Robins, J. (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21(1), C1–C68. (selected excerpts)

Bertsimas, D., & Kallus, N. (2020). From predictive to prescriptive analytics. Management Science, 66(3), 1025–1044.

Module 2: 
The course module literature includes part of:

Bernard W. Taylor. Introduction to Management Science, 12th edition, ISBN 978-0-13-377884-7, Pearson Education, 2016. Free e-book available.

Module 3:
The module will be based on articles and cases. The potential articles are as below (subject to change).

1. Ban, G.Y., Rudin, C., 2019. The big data newsvendor: Practical insights from machine learning. Oper-ations Research 67, 90–108.

2. Huber, J., Müller, S., Fleischmann, M., Stuckenschmidt, H., 2019. A data-driven newsvendor problem: From data to decision. European Journal of Operational Research 278, 904–915.

3. Dimitris Bertsimas, Nathan Kallus (2019) From Predictive to Prescriptive Analytics. Management Science 66(3):1025-1044. https://doi.org/10.1287/mnsc.2018.3253

4. Theodorou, E., Spiliotis, E., and Assimakopoulos, V. (2025). Forecast accuracy and inventory performance: Insights on their relationship from the m5 competition data. European Journal of Operational Research, 322(2):414–426.

Schedule spring 2027
TBA

Registration
Registration deadline: February 10th at noon: Registration form