Predict customer behaviour with Transformer(attention is all you need)


Please advice, am I thinking correctly: is it possible to represent customer behavior data from an online store as a sequence data? Because it is describing interactions of the customer with the shop through time.

So in this case N would be the number of users (or number of user sessions), T / time window I could set myself and D I could set also myself taking only event type (purchase, view, etc.) or something else like price, brand etc.(please see screenshot below) enter image description here

Please share your opinion Many thanks in advance

Julia Koncha

Posted 2020-12-05T11:24:21.987

Reputation: 3



Definitely, it is a good idea and has been attempted before. Have a look at Alibaba's paper which uses transformers:

There are also various other papers that use other types of seq2seq for recommender systems.


Posted 2020-12-05T11:24:21.987

Reputation: 56

thank you for links! – Julia Koncha – 2020-12-05T16:57:20.353

in this case, if I have have only e.g. 16 features, than my model dimension (d_model) will be it even possible? In the attention paper they have 512... – Julia Koncha – 2021-01-19T12:18:17.043