Member-only story
All your metrics are belong to us
Survival Analysis evaluation
Find the best model when predicting the occurrence of an event
You found yourself having to evaluate the likelihood of an event taking place in the future and most of your data on past events is incomplete. You decided to use or learn Survival Analysis methods and you got stuck at the model comparison stage. By the end of this article, you’ll have a good grasp on the model performance metrics and be able to pick that which works best of your use case.
Brief intro
Say you have found the house of your dreams: your soulhome. Though you are not ready to buy it, you have not considering faking poltergeists to get the current owners out their house. Instead you’re put together a financial plan so that you can snatch it off the market as soon as it appears there. The main ingredient in your plan is the time you have to get the funds together. To get this, you’ll have to predict when this event will take place. No problemo, you’re unafraid by Data Science. You gather public data on property sales to model length of ownership.
You soon realize- the majority of data available is that of two overlapping “populations”:
- sales for which the previous sale is so far in the past…
