Continuous WTTE-RNN case for churn prediction

[aprotopopov]

Is it possible to use WTTE-RNN when time is continuous or almost continuous? I have time scale as seconds and trying to predict next user session and not sure how to deal with TTE in that case.. Should it be every second? It'll be very high dimensional tensor and very sparse events. Could WTTE-RNN deal with such scenario? What about divide TTE between events on fixed number (e.g. between 5010 sec and 10 sec with 10 intervals will be 500 sec)? In that case it'll be different TTE step between each event. Now I'm thinking about switching to days but it's more natural to use sessions in my case because there could be about 20 user sessions per day.

[ragulpr]

Hi there, Sorry for the late answer but this is very much ongoing research 😅. I think what you are talking about is doing asynchronous updates whenever we get that, and that would be brilliant. I got the math and the model for it currently in my head but I've just yet to implement how that would be tested in a convincing way!

[aprotopopov]

Hi, Egil. Thanks for response. For now we decided to go with days. When you'll update your library/docs with continuous case we definitely will give it a try.

[ragulpr]

Thanks! Please keep posting questions and good luck!

[shalberd]

Would it be possible to use PhasedLSTM instead of LSTM for such events in continuous time? https://arxiv.org/abs/1610.09513 https://pypi.org/project/phased-lstm-keras/