Survival analysis is a set of statistical techniques that has many applications in the industry. This talk will discuss key concepts behind survival analysis by means of examples implemented via Lifelines, an open source python library, and in R for comparison purposes. I will also describe how we have made use of these techniques in Lyst to try to predict when items go out of stock.
Many problems involve the understanding the duration of specific events; for example, predicting when a customer will churn, when a person will default on a credit, how long a machine will work, etc. These type of questions constitute the realm of Survival analysis, a branch of statistics historically developed by professionals in the actuarial and medical fields dealing with event durations as governed by probability laws.
In this talk I will cover the basics of Survival analysis via examples implemented via Lifelines, an open-source python library and in R (survival and KMsurv libraries), going from survival curves to regression models. I will discuss how survival analysis can be applied to a variety of problems and in particular, I will focus on the problem of out of stock prediction for an online retailer.