Data science modelling

SOST70062 Data science modelling 15 Semester 2 Postgraduate Taught Level 7 Tatjana Kecojevic Unit coordinator 1 FHEQ level (Framework for Higher Education Qualifications) ' Masters/Integrated Masters P4 ' European Credit Transfer & Accumulation System Rating : 7.5 This is an introductory course Machine/Statistical learning, aimed at students interested in methods and models for large complex datasets comprising of numerous variables measured on different scales. The course will benefit students with an interest in quantitative Social Sciences (including Criminology, Politics, Psychology, or Sociology to mention but a few). As an introductory course, the focus will be on the underlying ideas concerning specific methods and models, as opposed to formula and/or theoretical results. Yet, we will emphasise that Statistical Learning should not be seen as a series of black boxes. The course content will be presented and practiced from a problem-oriented perspective using applications from the Social Sciences  This is an introductory course Machine/Statistical learning, aimed at students interested in methods and models for large complex datasets comprising of numerous variables measured on different scales. The course will benefit students with an interest in quantitative Social Sciences (including Criminology, Politics, Psychology, or Sociology to mention but a few). As an introductory course, the focus will be on the underlying ideas concerning specific methods and models, as opposed to formula and/or theoretical results. Yet, we will emphasise that Statistical Learning should not be seen as a series of black boxes. The course content will be presented and practiced from a problem-oriented perspective using applications from the Social Sciences  The unit aims to provide students with an understanding of how to handle high dimensional and complex data sets.To enable students to implement a battery of machine/statistical learning methods and models to address different classification and forecasting / prediction problems (both in supervised and unsupervised settings)To give students an assertive command of the statistical package R  Students will gain a number of skills that will be useful in academia (if pursuing a PH.D. later in their careers) and outside academia (in both, the public and private sectors). The key skills students will acquire are:    <ul><li>Select, among a pool of competing analytical tools, those most appropriate to the specific application   </li><li>Implement the selected tool using R   </li><li>Successfully write a report explaining and supporting the findings of their analyses and critically assess the results obtained  </li><li>Critically assess the methods being used   </li><li>Given a clear policy or research question, students will be able to undertake their own projects involving large and complex datasets </li></ul> <ul><li>Show critical awareness of the mathematical and statistical characteristics of the main classification and prediction methods for machine learning  </li><li>Apply methods for prediction and classification using free software (R or Python)</li><li>Be able to synthetise and explain the theoretical foundations of different model selection methods within a predictive or classification task (such as cross-validation)</li></ul> <ul><li>Make judgements about the expected performance of methods for prediction and classification in the face of complex data</li><li>Apply cross-validation  and other model selection methods  using free software (R or Python)</li><li>To model neural network models, balancing the complexity and predictive properties of the model, and apply these and other machine learning methods in a wide class of social research problems</li></ul> <ul><li>Apply prediction and classification methods using the statistical package R </li><li>Design and deploy neural network models using R </li><li>Write a report in Markdown, producing high quality tables of statistical results and visualisations using R</li></ul> <ul><li>Estimation and selection of models using the statistical package R </li><li>Write up a report to explain the conclusions of a statistical analysis </li><li>Produce tables and visualisations of statistical results</li></ul> Regression (General and generalised linear models) Discrimination and classification (e.g. logistic and nonparametric regression, tree-based methods) Regularisation Dimension reduction. Unsupervised learning Neural networks   The unit will be based on eleven two-hour lectures, spread across the second term, as described in section 1. There will be ten one-hour tutorials to support the lectures. Students will also be given 3 online quizzes to evaluate their understanding of the materials. 0 Other 100% Four online quizzes: 4 x 10%Final written report: 2,000 words, 60% Formative assessment feedback will be provided during the live lectures.Summative assessment feedback: students will receive their mark upon completion of online quizzes, and feedback will be provided through Turnitin within 2 weeks of submission of the final written report. SOST70151 Statistical Foundations Pre-Requisite Compulsory N The unit will be based on Chapters 1 to 10 of   James, G., Witten, D., Hastie, T., Tibshirani (2017). An Introduction to Statistical Learning: with Applications in R. Springer.  James, G., Witten, D., Hastie, T., Tibshirani (2023). An Introduction to Statistical Learning: with Applications in Python. Springer.  Additional readings from popular press and published research will be provided in an ad hoc basis, to support each week’s class activity.     Lectures 22 Seminars 8 0 120