i) the use of systematic and reproducible approaches to high-quality predictive analytics projects, such as the CRoss-Industry Standard Process for Data Mining (CRISP-DM); and
ii) the exploitation of the latest discoveries in machine learning technologies, such as artificial neural networks (deep recursive autoencoder nets, convolutional nets, residual nets, long short-term memory nets, self-organising maps), decision tree learning models (xgboost, catboost, and random forests), Bayesian methods and support vector machines. Our emphasis is primarily on first establishing the organisations goals and objectives and their current business processes, technical infrastructure and employee talent for achieving those objectives. We then move through the CRISP-DM methodology, in manner suited to the business context and project requirements, to work collaboratively with the workforce to understand, visualise, transform and model their data until the resulting solution meets the desired objectives…