CdSDATA SCIENCE AND BUSINESS INFORMATICS
By the end of the course:
- Students will have acquired knowledge about the online advertising industry, both of economic and technical aspects.
- Students will have acquired general knowledge about decision making under uncertainty. This kind of knowledge will be also applicable in other sectors, mainly e-commerce and finance (investments).
- Students will be able to demonstrate knowledge required to enter professional teams working on digital advertising.
- Studnets willl have acquired a "mindset" and a conceptual toolkit enabling them to correctly frame certain kind of problems, with awareness of foundations, merits and limits of several methodologies.
During lessons the teacher proposes problems to be discussed, driving the debate towards certain lines of reasoning. Students can assess their progress depending on their capability to correctly frame the problem (not necessarily solving in detail) and their awareness of the mindset required.
During exam, the same criterion will be applied: students will demonstrate they are able to master a repertoire of concepts and methods.
Students will be able to:
- Analyze the performance of online dvertising campaigns.
- Make educated decision about campaign management.
- Design algorithms for camapign performance optimizatipn.
Some "open" problems will be submitted, i.e. problems without a precise pre-defined and unique solution.
During lessons students can assess their capability to design a solution scheme.
The teacher will do the same assessment during the exam.
Students will be able to join a team in the online advertising industry and quickly contribute with ideas and methods to data analysis, optimization of outcomes, and automatization of workflows.
Both during lessons and during exams, students will discuss problems which typical of online advertising and advance proposals to empower optimization and automatization.
Basics of probability are useful, though not mandatory.
Digital advertising industry.
Online campaigns management, reporting, optimization.
Decision making, uncertainty and risk, utility maximization, exploration-exploitation problem.
Bayesian approach to decision making.
Multi-armed bandits problems and methods: Epsilon-Greedy, Softmax, Thompson Sampling, Upper Confidence Bound.
Predictive marketing problems and methods: bayesian predictors, linear and logistic regression , applcation to programmatic advertising.
Dynamic programming methodology.
Appllications to e-commerce. extension to revenue management and dynamic pricing.
Lecture notes are sufficient.
Some additional resources will be proposed, though not required for the exam.
Non-attending students are invited to contact the teacher and arrange an appointment.
This is very useful in order to understand the very spirit of the course and make thestudy mpre proficient.
The exam consists in an oral discussion.
Initially an open question will invite the student to speak about a substantial topic, e.g. "Let's speak about multi-armed bandits". In this phase the goal is testing whetherthe student is able to frame a certain topic and related problems. This is the necessary requisite to succesfully pass the exam.
Afterwards, deep questions will be submitted, e.g. "What if we have a problem different from the standard one? How can we adapt and tune our methodology to cope with it?". In this phase teh goal is testing whether the student is able to generalize principles and methods beyond teh domain explicitly discussed in lessons and notes. This is the criterion driving the exam mark.
It is strongly suggested that students spend their study effort on principles and methods more than in technical details.