Duration
30h Th
Number of credits
Lecturer
Language(s) of instruction
English language
Organisation and examination
Teaching in the second semester
Schedule
Units courses prerequisite and corequisite
Prerequisite or corequisite units are presented within each program
Learning unit contents
The course presents various methods and tools of artificial intelligence and machine learning applied to business analytics problems, with an emphasis on their understanding, implementation, and critical evaluation.
The main topics covered are:
- Anomaly detection: methods such as Support Vector Data Description (SVDD) and One-Class SVM, possibly applied to the residuals of other models to detect atypical behaviours, process failures, or potentially fraudulent situations.
- Autoencoders and Variational Autoencoders (VAEs): learning representations and generating data, in particular for simulating alternative economic or financial scenarios.
- Generative models: Generative Adversarial Networks (GANs) and diffusion models, in particular for generating and sharing synthetic or anonymised data.
- Transformers and language models: BERT- and GPT-type architectures, representation and classification of textual data, embeddings and clustering, as well as applications to forecasting and other analytical problems in business.
- Artificial intelligence platforms and tools: critical use of AI platforms and AutoML tools for implementing business analytics solutions.
- Model interpretability and explainability: post-hoc methods such as SHAP and LIME, considered in relation to intrinsically interpretable models and to the importance of interpreting results in a managerial decision-making context.
Learning outcomes of the learning unit
By the end of the course, students will be able to critically use data, statistical methods, and artificial intelligence tools to address business analytics problems and support decision-making.
More specifically, students will be able to:
- formulate a business problem as a problem that can be addressed through a data analytics or artificial intelligence approach;
- identify, select, and use relevant data according to the objective pursued, taking into account data quality, limitations, and the context in which the data were generated;
- identify and select appropriate analytical methods and tools among different statistical, machine learning, and artificial intelligence approaches, depending on the nature of the problem and the available data;
- implement and evaluate analytical and machine learning methods on real or simulated data;
- rigorously analyse and interpret results, assessing in particular their relevance, robustness, limitations, and associated uncertainty;
- demonstrate critical thinking with regard to data, models, and results, in particular by identifying biases, assumptions, overfitting, and limitations to generalisation;
- assess the interpretability of models and results, and distinguish between intrinsically interpretable approaches and post-hoc explainability methods;
- visualise and effectively communicate results from an analysis, adapting their presentation to the context and needs of different audiences, including decision-makers;
- critically assess the use of AI solutions in a business context, relating their technical performance to their relevance and value for the business problem under consideration.
Prerequisite knowledge and skills
Foundations in constrained optimization, probability, and statistical inference. Equivalent to the content covered in the following courses: MATH0059-2 Mathematics for Business Engineers and STAT1208-2 Probability and Statistical Inference.
Planned learning activities and teaching methods
Blended learning
Further information:
Used methodology
A3. Analysis of a practical problem by each group of students (partially followed up by the teacher).
A4. Critical synthesis of searches, readings and/or practical applications achieved by each group of students. Each student presents his own results and then, each group discusses, compares the different methods and presents other possible obtained results.
During his talk, each group is invited to
1) clearly present the problem of interest in its context and the existing methods to solve it,
2) discuss those methods and justify the choice of one or several of them in specific cases.
Moreover, each student is expected to attend to presentations of the other students and discuss the way they treat their own problem.
Overview of the course agenda
The first week, the teacher presents the different problems of interest with the necessary corresponding theoretical basic knowledge. Then, students (in groups) receive a problem and try to understand it, on one side, with a personal bibliographic search (in agreement with their group -shared bibliographic search-) and on the other side, with the courses and/or discussions conducted the weeks after. The sessions are designed to both follow up each group in his work and provide a general understanding of the work achieved by the other groups. After some weeks, the students and the teacher meet to assess the progress of the work and define the remaining steps to achieve. Finally, the students prepare an oral presentation of their problem and write a report for the evaluations period that follows the course.
Decomposition of the student workload
A1 Lectures 24h
A3 Analysis of the problem 60h
A3 State of the progress, meetings with the teacher 10h
A4 Report 20h
A4 Presentation 6h