Supervised and unsupervised learning, neural networks, support-vector machines, decision trees, metric-based clustering, distribution-based clustering, rule-based techniques, genetic algorithms. Applications to information retrieval, web mining, customer-relationship management, recommender systems, science and engineering. The main objective of this course is ensure that students know enough about the algorithms, strengths and limitations of mainstream data-mining techniques that they can use data-mining software appropriately, and can understand the results that are produced. In particular, they should be able to see how to model a real-world problem, choose appropriate algorithms, analyse the results, and explain their implications for the original problem. A smaller objective is to make students aware that not all problems in computing have a single cut-and-dried, correct solution. A major component is a 6-week design project in which students are given a real-world dataset, and are asked to solve an open-ended data-mining problem related to it.
