WTTE-RNN - solving churn prediction by predicting time to event

We will look at the deceptively complex problem of churn prediction and use it as a case for trying out a combination of machine learning and more traditional statistics. If you don't know what churn prediction is, don't worry, because most people don't. We will circumvent this problem by looking at the expected Time To some Event (TTE). For instance trying to predict the time to next purchase might be better than trying to predict the chance of a customer never purchasing something ever again.

The talk will explain what a Weibull distribution is, and how a Recurrent Neural Network (RNN) may predict such a distribution.

Based on https://ragulpr.github.io/2016/12/22/WTTE-RNN-Hackless-churn-modeling/

About the speaker: Ask is a Data Scientist at AIA Science. He holds a Master's degree in Applied Physics and Mathematics. He is known to be an all around average guy, but possesses a beard some would describe as longer than average.

Intended audience: People who are interested in machine learning or statistics. The talk will explain some mathematics.

As usual, there will be pizza and soft drinks :) A big thanks goes to Den Norske Dataforening (DND) for sponsoring this!