About this course
Shows how to use machine learning to predict business outcomes and act on the predictions. Applications include credit scoring, customer churn, fraud detection, demand forecasting and revenue collection risk. Delegates cover target definition, feature engineering, classification and regression models, imbalanced data, model validation, interpretation with feature importance and the steps for putting a model into regular use.
Why this course
Lenders, utilities and revenue authorities make thousands of decisions that depend on predicting who will default, leave, commit fraud or fail to pay. Rule-of-thumb approaches miss patterns and treat everyone the same. Well-validated predictive models let organisations target collections, inspections and retention efforts where they matter most, while keeping decisions explainable to auditors and regulators.
What delegates will be able to do
By the end of the course, delegates will be able to
- Define a prediction target and success measure linked to a business decision
- Engineer predictive features from transaction, customer or operational data
- Build classification models and handle imbalanced classes in fraud or default data
- Validate models using holdout samples, ROC curves and lift charts
- Prepare a deployment and monitoring plan for a predictive model
Who should attend
Credit risk analysts, fraud and revenue analysts, actuarial and marketing analysts and BI specialists who want to apply predictive models to operational decisions. Delegates should have basic statistics and some experience with Python or a similar analysis tool.
How it is delivered
Case-based practical sessions in Python on credit, churn and fraud datasets, with group presentations translating model results into business recommendations. It is available as a public course, in-house at your premises, or as live virtual sessions. Duration and dates are confirmed in your quote, and in-house groups can have the examples and exercises fitted to their organisation.