Fundamentals of Predictive Modelling

In each workshop session, you will encounter a real-life public sector dataset. The trainer will guide you through the data, exploring and learning how to think critically about how to manipulate and model it.

Workshop 1 focusses on R and preparing data, workshops 2 and 3 on statistical models, workshop 4 on machine learning models, and workshop 5 on implementing a predictive model within the organisation, including time for discussion specific to attendees’ settings.

• Using R on Windows or Mac: installation and basics of the language
• Loading and manipulating data using the Tidyverse packages for R
• Examining data visually using the ggplot2 package in R
• Proposing and justifying a predictive model

• Understand how regression models are made up of three key assumptions about the data
• Use R code to fit a linear regression model to the data
• Gain confidence in interpreting the outputs
• Visualise the model and compare it critically to the data
• Identify problems in the data that may undermine a predictive model
• Extend linear regression into more than one predictor variable

• Use the LASSO algorithm to select predictor variables from a large collection
• Extend linear regression into predicting binary variables with logistic regression using R
• Interpret and visualise the results
• Communicate risk effectively to non-technical audiences
• Understand how linear and logistic regression are specific instances of generalised linear models (GLMs)

• Use R to fit a decision tree to data
• Understand the choices in using decision and regression trees, such as pruning
• Compare different ways of communicating the predictions
• Extend into an averaged prediction across an ensemble of models: random forest
• Know the differences between statistical and machine learning models, and their pros and cons

• Identify your key service delivery objectives and map these to predictive model strengths and weaknesses
• Understand how predictive models can be validated and revised on an ongoing basis
• Discuss enablers and barriers within a variety of public sector organisations
• Examine different approaches to embedding the data analytic workforce