Course content

The course is an introduction to chemometric methods and data analysis with emphasis on applications for chemistry, biotechnology, process chemistry, material science, and physics.

The course covers methods for the design of experiments, pre-processing, and modelling of measured data to extract useful information from possibly large data sets, and use this for supporting decisions. Specifically, the following themes are covered:

• simple regression (e.g. least squares and polynomial regression),
• experimental design (full and fractional factorial design),
• pre-processing (e.g. auto-scaling, Fourier filtering, Savitsky-Golay filtering and numerical differentiation, convolution),
• reduction of large data sets to interpretable information, for instance, via methods such as principal component analysis, principal component regression, and partial least squares regression,
• validation of models (by the use of test sets, cross-validation, bootstrap, and y-randomisation),
• cluster analysis (e.g. hierarchical and k-means cluster analysis),
• classification (e.g. random forest and k-nearest neighbors),
• introduction to machine learning techniques for classification, regression, and clustering.

Learning outcome

Knowledge

After completing the course, the student can:

• Explain the difference between supervised and unsupervised methods and select if a supervised or unsupervised method is the most appropriate for different situations.
• Explain how experimental design is used for planning experiments and how the results from such designs are analysed.
• Give examples of different pre-processing methods and select the most appropriate method in different situations.
• Describe unsupervised methods such as principal component analysis and clustering methods, give examples of their use, interpret and assess results from such methods.
• Describe regression methods (e.g., least squares and partial least squares), give examples of their use, interpret and assess results from such methods.
• Describe classification methods, give examples of their use, interpret and assess results from such methods.
• Explain how validation methods are used for assessing the predictive ability of different models.
• Indicate the limits of the different methods and models covered in the course.

Skills

After completing the course, the student can:

• Reduce and simplify large datasets to interpretable information.
• Set up, carry out, and interpret results from experimental designs.
• Carry out pre-processing for different situations.
• Carry out principal component analysis and cluster analysis, and use these methods for interpreting large data sets.
• Carry out regression and use this for modelling and prediction.
• Carry out classification and use this for modelling and prediction.
• Make use of test sets and cross-validation for describing and comparing the predictive ability of different models.

General knowledge

After completing the course, the student can:

• Present results from modelling and analysis in written and graphical form.
• Make use of Python for simple scientific analysis and plotting, in particular, for the different methods covered in this course.

Learning methods and activities

• Lectures.
• Exercises.

A certain number of the exercises (specific details will be given at the start of the course) have to be completed and approved in order to take the written exam.

Information about the start of lectures and compulsory activities will be given via Blackboard.

Expected work load in the course is 200-225 hours.

Compulsory assignments

• Exercises

Recommended previous knowledge

Basic level in mathematics (in particular linear algebra), statistics and chemistry.

Course materials

The course material will be announced at the beginning of the course.

Credit reductions