The unsupervised “2vec” algorithms (e.g. word2vec) learn embeddings extremely quickly, scale well and can create beautiful visualisations. dunnhumby have recently extended these algorithms to work with grocery retail data; helping us to represent products, baskets and customers within the same multidimensional space. In this talk, I’ll explain how this has helped us to understand customer preferences better and discuss some of the potential use cases.
Supported by
