An Approach to Analytics Modelling in the ATO by Dr Graham Williams, Corporate Analytics, The Australian Taxation Office
On Thursday 26 February, Dr Williams presented an overview of the use of ensembles in Data Mining to the South Australian analaytics community. The presentation introduced the concepts of ensembles, big data and ensembles for big data in R in the ATO. The presentations provide an overview of their approach; how they developed the concept of multiple models, found it useful, and developed it into boosted decision stumps, random forests, and ensembles of nuggets.
For context, the ensemble concept was presented at the first Australian AI Conference in Sydney in November 1987, with Ross Quinlan as the session chair. As recounted in a chapter in Journeys to Data Mining (Mohamed Medhat Gaber), the idea of not going with the best single model, but combining multiple models was challenged – it is now the approach of choice for many data scientists.
You are welcome to download a copy of the presenatation for your own reference. We look forward to seeing you at future events.