Mercator Endowed Chair

This course is dedicated to conveying a sense of how analytics projects work so as to be able to manage them and/or assess their merit. 

It is not a modelling course - although we will do modelling. It is also not a programming course - although we will do plenty of programming in R. Instead, the modelling and programming just serves as an illustration of the steps featured in typical analytics projects. This should help in the planning of such a project, starting from understanding of the business problem over modelling up to model assessment and communication of the project's results (or a project proposal) to a client.

There is no classic split between lecture sessions and tutorial sessions; instead, lecture elements, practical demonstrations and exercises are mixed together in all sessions so as to create a more engaging environment. In an assessed groupwork, you will go through all the stages of a data science project including shaping the business objectives and connecting the modelling results to them.

We will also cover visualization concepts in both theory and practice, using Tableau for the latter. In particular, we will look into dashboard design, interactive maps (such as the one shown in Fig 1) and charts, and how to structure sales pitches.

The syllabus looks as follows:

1. Introduction to the CRISP-DM process (business understanding)
2. Sampling and Partitioning (data preparation)
3. Information selection, modelling and overfitting (modelling)
4. Model evaluation
5. Evidence combination (Naïve Bayes, association mining) and visualization
6. Visualization, dashboards, selling your project to end users

This course is dedicated to conveying a sense of how analytics projects work so as to be able to manage them and/or assess their merit. 

It is not a modelling course - although we will do modelling. It is also not a programming course - although we will do plenty of programming in R. Instead, the modelling and programming just serves as an illustration of the steps featured in typical analytics projects. This should help in the planning of such a project, starting from understanding of the business problem over modelling up to model assessment and communication of the project's results (or a project proposal) to a client.

There is no classic split between lecture sessions and tutorial sessions; instead, lecture elements, practical demonstrations and exercises are mixed together in all sessions so as to create a more engaging environment. In an assessed groupwork, you will go through all the stages of a data science project including shaping the business objectives and connecting the modelling results to them.

We will also cover visualization concepts in both theory and practice, using Tableau for the latter. In particular, we will look into dashboard design (and create a few such as the one in Fig. 1), interactive maps and charts, and how to structure sales pitches.

The syllabus looks as follows:

1. Introduction to the CRISP-DM process (business understanding)
2. Sampling and Partitioning (data preparation)
3. Information selection, modelling and overfitting (modelling)
4. Model evaluation
5. Evidence combination (Naïve Bayes, association mining) and visualization
6. Visualization, dashboards, selling your project to end users
7. Tableau: using web data connectors, calling R from within Tableau, and other more advanced topics

Pricing analytics and revenue management focuses on how a firm should model demand, set and update automated pricing and product availability decisions across its various selling channels in order to maximize its profitability. The use of such strategies has transformed the transportation and hospitality industries, and they are increasingly important in retail, telecommunications, entertainment, financial services, health care and manufacturing.

Within the broader area of pricing theory, the course places emphasis on tactical optimization of pricing and capacity allocation decisions, tackled using demand modeling and constrained optimization – the two main building blocks of revenue management systems.

Case studies provide hands-on experience of the subject. Students are using R for most of the exercises within the RStudio environment, involving training on both demand modeling and optimization problems. For example, in the context of B2B customized pricing, we look into the question of how to estimate the win probability function from historical data and how to use this to optimize individual price quotes.

The syllabus consists of the following:

1. Introduction, customer valuation game
2. Demand modelling (parametric, non-parametric models, unconstraining)
3. Constrained price optimization, capacity control, network revenue management
4. Dynamic price control, (approximate) dynamic programming
5. Markdown pricing, behavioural pricing
6. Customized B2B pricing, win probability function estimation

Optimization is important to many applications in business, be that finance, operations, marketing or others. This course aims to provide a broad overview of the concepts that underpin optimization to help students to gain an understanding of what type of optimization problem they may be dealing with in their studies, and how this could be tackled.

Coverage includes:

• Structure of an optimization problem
• Deterministic versus stochastic optimization
• Continuous versus discrete optimization
• Constrained versus unconstrained optimization
• Fundamentally important concepts like convexity, duality, complexity, total unimodularity, ...
• Introduction to various techniques including linear and non-linear mathematical programming, (approximate) dynamic programming for control problems, optimal learning

We will not go overly deep into the topics due to time constraints; instead, the focus is on imparting an intuitive understanding of optimization techniques and of structures that can be exploited. The intention is to make this course useful and relevant to any students who face some form of optimization problem and who do not yet have received formal training in optimization.