¹û¶³Ó°Ôº

The context for the Predictive Analytics module is management in complex, interconnected, data-driven environments.

Predicting or Forecasting is a fundamental business skill. Forecasts of the future are used in all areas of business, from operations and finance to marketing and entrepreneurship. Predictive analytics is about using data to forecast uncertain quantities and events.

Predictive Analytics introduces students to the main ideas behind approaches to prediction spanning Applied Artificial Intelligence, Machine Learning and Data Science.

The first part of the course focuses on the traditional Data Science process from start to finish where we introduced the four main Machine Learning tasks:

• Regression
• °ä±ô²¹²õ²õ¾±´Ú¾±³¦²¹³Ù¾±´Ç²ÔÌý
• Clustering; andÌý
• ¶Ù±ð³¦´Ç³¾±è´Ç²õ¾±³Ù¾±´Ç²ÔÌý

The main algorithms that will be reviewed include: Lasso and Ridge Regression, Logistic Regression, Classification and Regression Trees, K-means, Mixtures of Gaussians and Principal Components Analysis. The main optimisation algorithms that are introduced include Gradient Descent, Hierarchical greedy optimisation, and Expectation Maximisation. We also look at how to combine models together using Ensemble methods, for example Boosting and Stacking.

The second part of the course focuses on modern advances in Representation Learning including Neural Networks and Deep Learning. We will introduce the concepts of shallow and deep networked models and then look at several different specialist architectures, including:

• Convolutional Neural NetworksÌý(for images)
• TransformersÌý(for text)
• Recurrent Neural NetworksÌý(for timeseries); and
• ´¡³Ü³Ù´Ç±ð²Ô³¦´Ç»å±ð°ù²õÌý

The module builds on ideas and tools introduced in MSIN0096 Statistical Foundations of Business Analytics and MSIN0143 Programming for Business Analytics, including linear algebra and the use of the Python programming language. New programming modules that are introduced include Scikit-Learn, Tensorflow, PyTorch and XGBoost. It also includes access to various state-of-the-art API ML services.

During the module, students will work with example data sets to experience the stages of the data science process: they will visualise data, propose models that might fit the data, choose a best-fit model, use that model to make predictions, and test those predictions against new realisations.

The coursework, both individual and group submissions, develop students’ ability to think and work like a data scientist and apply both classical and modern techniques.

The aims of the Predictive Analytics module are:

• To develop and understanding of the role of data in the modelling processÌýÌý
• To develop a rigorous understanding of how machine learning is used to support the practice of management and strong data-based reasoning and computational thinking skills.
• To understand the main components of a Machine Learning system: The choice of function family, Objective function, Optimizer, Regularizer and the test harness for training.
• To introduce students to modern machine learning techniques including Deep Learning, Generative Pre-trained Transformer models and Prompt Engineering.