The main focus of the course will be supervised learning, primarily for classification. The emphasis will be on practical applications of the methodologies that are described, with the R system used for the computations. Attention will be given to

1) Generalisability and predictive accuracy, in the practical contexts in which methods are applied.

2) Low-dimensional visual representation of results, as an aid to diagnosis and insight.

3) Interpretability of model parameters, including potential for misinterpretation.

There will be very limited attention to regression methods with a continuous outcome variable. Relevant statistical theory will mostly be assumed and described rather than derived mathematically. There will be somewhat more attention to the mathematical derivation and description of algorithms.

Topic to be covered include:

• Basic statistical ideas - populations, distributions, samples and random samples
• Classification models and methods - including: linear discriminant analysis; trees; random forests; neural nets; boosting and bagging approaches; support vector machines.
• Linear regression approaches to classification, compared with linear discriminant analysis,
• The training/test approach to assessing accuracy, and cross-validation.
• Strategies in the (common) situation where source and target population differ, typically in time but in other respects also.
• Unsupervised models - kmeans, association rules, hierarchical clustering, model based clusters.
• Low-dimensional views of classification results - distance methods and ordination.
• Strategies for working with large data sets.
• Practical approaches to classification with real life data sets, using different methods to gain different insights into presentation.
• Privacy and security.
• Use of the R system for handling the calculations.

Note: This is an HPC, available as an HPC for students with outstanding results in mathematical and/or computing later year courses. Students will be required to do an indepth presentation of a current research topic, as well as demonstrate the use of advanced data mining techniques on data sets from numerous application areas.

On satisfying the requirements of this course, students will have the knowledge and skills to:

Assessment will be based on:

• 3 Assignments (60%; LO 1-5)
• Presentation (30%; LO1-5)
• Commentary on other Presentations (10%; LO 1-5)