English   Danish

2024/2025  KAN-CDSCO1005U  Predictive Analytics

English Title
Predictive Analytics

Course information
Language English
Course ECTS 7.5 ECTS
Type Mandatory (also offered as elective)
Level Full Degree Master
Duration One Semester
Start time of the course Spring
Timetable Course schedule will be posted at calendar.cbs.dk
Study board
Master of Science (MSc) in Business Administration and Data Science
Course coordinator
  • Herdis Steingrimsdottir - Department of Economics (ECON)
The teacher need not be the course responsible person.
Main academic disciplines
  • Statistics and quantitative methods
  • Economics
Teaching methods
  • Blended learning
Last updated on 12-11-2024

Relevant links
Learning objectives
To achieve the grade of 12, students should meet the following learning objectives only with no or minor mistakes or errors. By the end of the course the students will be able to:
  • Explain the concepts, models and methods introduced during the course.
  • Identify and perform an academically founded forecasting analysis in practice (including taking relevant theoretical relationships into account) using the tools introduced during the course.
  • Discuss and solve any important problems encountered in relation to the analysis.
  • Evaluate the forecasts of the analysis.
  • Report/communicate the results and conclusion of the analysis both if the reader is a statistician/econometrician and if the reader is e.g. a CEO.
  • Evaluate a forecasting analysis conducted by another person/researcher.
  • Use a statistical analysis software package for the analysis (e.g. R or SAS) and demonstrate that you can interpret and evaluate the output from the software package.
Course prerequisites
The course assumes familiarity with either a basic courses in statistics and regression analysis or a basic course in econometrics. Concepts such as random variables and familiarity with the principles of statistical hypotheses testing are assumed known.
Prerequisites for registering for the exam (activities during the teaching period)
Number of compulsory activities which must be approved (see section 13 of the Programme Regulations): 2
Compulsory home assignments
Each assignment is 3-5 pages in group of 1-4 students.
The students have to get 2 out of 3 assignments approved in order to go to the exam.

There will not be any extra attempts provided to the students before the ordinary exam.
If a student cannot hand in due to documented illness, or if a student does not get the activity approved in spite of making a real attempt, then the student will be given one extra attempt before the re-exam. Before the re-exam, there will be one home assignment (max. 10 pages) which will cover 2 mandatory assignments.
Examination
Predictive Analytics:
Exam ECTS 7,5
Examination form Home assignment - written product
Individual or group exam Individual exam
Size of written product Max. 10 pages
Assignment type Project
Release of assignment Subject chosen by students themselves, see guidelines if any
Duration Written product to be submitted on specified date and time.
Grading scale 7-point grading scale
Examiner(s) One internal examiner
Exam period Summer
Make-up exam/re-exam
Same examination form as the ordinary exam
Students can submit a revised project.
Description of the exam procedure

The paper is a reporting of an analysis using R and applying at least 2 of the time series model types given in the syllabus.  Data and topic are chosen by the individual student. No appendices are allowed except for documentation of R code and data (could be a knitted R file). This documentation will not be part of the assessment. A document uploaded to CANVAS will provide more details on potential content and structure. In principle the students can start writing on their projects from the date where the course starts. However in most cases the students will wait until they know more about the content of the syllabus. The project is individual and must relate in a satisfactory way to all the learning objectives of the course in order to lead to the top grade.
Course content, structure and pedagogical approach

This course is designed to provide both a theoretical foundation for predictive modeling but also hand-on experience with analyzing economic data. The models presented in the course will belong to different fields of statistics but in all cases the predictive power of the models will be in focus. Introducing statistical tools allows us to not only come up with an actual forecast of an economic variable of interest (think e.g. of sales) but also to assess the uncertainty of the forecast. Before a proper model can be selected it is of interest to discuss issues about data availability and also data reliability. There is often negligible discussion of such topics in the text book but providing the students with examples of the types of challenges one may encounter – based on the research of the teachers – will try to fill this gap. 

 

 

Predictive analytics is a field that can provide managers with very valuable tools for decision making. Both for sales variables of firms but also for financial variables like stock prices forecasting models and methods may be of interest, and during the course we will discuss challenges in relation to preparing relevant data for forecasting, choosing well-suited models for analysis and we offer a range of tools for evaluating the forecasting performance of the models we present. In one of the lectures we introduce a tool that can be very helpful for model selection if ‘Big  Data’ is available.

 

The course contributes to several competences emphasized at CBS, such as emphasizing critical thinking, providing students with deep business knowledge in broad context, and being encourage students to be analytical with data and curious about ambiguity. 

 

 
Description of the teaching methods
Lectures, in-class exercises during lectures, exercise classes with tools training.
Feedback during the teaching period
The main feed-back element is the feed-back provided for the mandatory assignments. This feedback consists of an indicative solution uploaded to Canvas followed by comments in class. Also individual feed-back is provided as comments in the individual assignments are made available to the students and students are encouraged to schedule a talk with the teacher about mistakes.

Next the teacher will create discussion fora on CANVAS where students can ask question and the answers will be displayed to the benefit of the entire class. This is a second way of receiving individual feedback.

Also feed-back is available more generally on a continuous basis in class. This holds true for both lectures and exercise classes. In addition the students are very welcome during the office hours of the teacher. During office hours students can receive feed-back both on their choice of data for the exam project, on technical issues in relation to the use of R in general (only to a limited degree for their exam project as no supervision is allowed. Minor questions can be answered if the teacher does not find it unfair to the other students). Also feed-back on theoretical questions is provided during office hours.
Student workload
Class lectures 24 hours
Exercise classes 24 hours
Class preparation incl. home assignments 100 hours
Exam and exam preparation 58 hours
Total 206 hours
Expected literature

The literature can be changed before the semester starts. Students are advised to find the final literature on Canvas before they buy the books.

 

Main text book(s) and articles:

Hyndman, R.J & Athana­sopou­los, G (2021): Forecasting: principles and Practice   (HA) (Free online version available)

 

Gujarati, D.N. and Porter, D.C. (2010). Essentials of Econometrics, 4th edition, McGraw-Hill, pp 201-205 and 386-397

 

Doornik, J.A. & Hendry,  D.F.  (2014): Statistical model selection with ‘Big Data’. University of Oxford, Department of Economics, Discussion Paper Series,  # 735  (or more recent text in this spirit). (DH)

 

More supplementary stuff:

Buus Lassen, N., la Cour, L., Vatrapu, R. (2017), ‘Predictive Analytics with Social Media data’ in Sloan & Quan-Haase ed. The SAGE Handbook of Social Media Research Methods, Chapter 20, pp 328-341 (BCV)

 

Pretis, F., Reade J.J. and Sucarrat G. (2018). Automated General-to-Specific (GETS) Regression Modeling and Indicator Saturation for Outliers and Structural Breaks, Journal of Statistical Softeware, vol 86, 3, pp 1-44

 
Last updated on 12-11-2024