TK8116 - Multivariate Data and Meta Modelling: Preparing for Big Data Cybernetics


Examination arrangement

Examination arrangement: Oral examination
Grade: Passed/Failed

Evaluation form Weighting Duration Examination aids Grade deviation
Oral examination 100/100 D

Course content

Preparing for Big Data Cybernetics: The need for understandable, realistic modelling to deal with Quantitative Big Data
Different science traditions: Induction, deduction and the “-ometrics” cultures
Personality types and choice of mathematical modelling style
“Soft multivariate data modelling” by subspace approximation and subspace regression (Dynametrics, chemometrics, qualimetrics): Machine learning, with fast algorithms, graphical interpretation and statistical validation. Bridging between “hard” mechanistic modelling and “soft” data-driven modelling, staying away from “black box” deep learning.
Principle Component Analysis (PCA), ICA, MSC: Analysis of one data table
Ongoing R&D: On-the-fly processing of “everlasting” high-dimensional data streams
Classification, discrimination and cluster analysis
Nonlinear preprocessing, nonlinear soft modelling
Multivariate instrument calibration and PLSR-based selectivity enhancement: “Math is cheaper than physics”
Model validation
Partial Least Squares Regression (PLSR): Analysis of two data tables
Cost-effective statistical design of laborious experiments and massive computer simulations
Multivariate metamodelling to convert mechanistic models (ODEs, PDEs, FEM etc) to bilinear model form, for faster computation, easier overview and safer parameter identification.

Learning outcome

Preparing for Big Data Cybernetics: How to discover the Real World
KNOWLEDGE: In-depth knowledge of techniques for multivariate analysis of data. Knowledge of data-driven modelling (metamodelling).
SKILLS: Be able to build models based on experimental data using the aforementioned methods.
GENERAL COMPETENCE: Skills in applying this knowledge and proficiency in new areas and complete advanced tasks and projects. Skills in communicating extensive independent work, and master the technical terms of multivariate analysis. Ability to contribute to innovative thinking and innovation processes.

Learning methods and activities

Lectures, mandatory group work and project.

Compulsory assignments
• Exercise

Compulsory assignments

  • Exercise

Further on evaluation

Specific conditions:
Exam registration requires that class registration is approved in the same semester. Compulsory activities from previous semester may be approved by the department.

Specific conditions

Exam registration requires that class registration is approved in the same semester. Compulsory activities from previous semester may be approved by the department.

Required previous knowledge

Desire for mental overview, statistical validity and search for unexpected patterns and causalities.

Good understanding of basic vector and matrix algebra.

Course materials

Martens H (2015) Quantitative Big Data: Where Chemometrics can contribute, J. Chemometrics 15; J. Chemometrics 2015; 29: 563–581 1
Kristin Tøndel and Harald Martens (2014) Analyzing complex mathematical model behavior by PLSR-based multivariate metamodeling. WIREs Computational Statistics, Volume 6, Issue 6, pages 440–475, November/December 2014. DOI: 10.1002/wics.1325
Harald Martens (2011): The informative converse paradox: Windows into the unknown. Chemometrics and Intelligent Laboratory Systems 107 (2011) 124–138
Raffaele Vitale, Anna Zhyrov, João F. Fortuna, Onno E. de Noord, Alberto Ferrer, Harald Martens (2017): On-The-Fly Processing of continuous high-dimensional data streams. Chemometrics and Intelligent Laboratory Systems 161 (2017) 118–129



Examination arrangement: Oral examination

Term Statuskode Evaluation form Weighting Examination aids Date Time Room *
Autumn ORD Oral examination 100/100 D
Spring ORD Oral examination 100/100 D
  • * The location (room) for a written examination is published 3 days before examination date.
If more than one room is listed, you will find your room at Studentweb.