Today Gavin presented some of our work at the American Marketing Association Winter Marketing Academic Conference in Las Vegas, USA.
The paper, entitled “Temporal Topic Modelling for Big Retail Data” (Dr G Smith, Dr J. Goulding, M. Iliffe and Prof. A. Smith) This research details a novel method for segmenting/summarizing consumers based on the automatic identification of a number of interpret able underlying temporal purchasing patterns (consumer missions). More information (and photos of the event) after the link.
The approach harnesses non-negative tensor factorization to uncover factors that correspond to underlying “shopping missions”. These can then be used as basic building blocks for analysis of each consumer’s purchasing behaviour (a person or a store can be accurately represented as a weighted expression of those factors). Details of the work are available as a project webpage.