Kolmogorov-Smirnov (KS) Statistics is one of the most important metrics used for validating predictive models. It is widely used in BFSI domain. If you are a part of risk or marketing analytics team working on project in banking, you must have heard of this metrics.
READ MORE »
What is KS Statistics?
It stands for Kolmogorov–Smirnov which is named after Andrey Kolmogorov and Nikolai Smirnov. It compares the two cumulative distributions and returns the maximum difference between them. It is a non-parametric test which means you don't need to test any assumption related to the distribution of data. In KS Test, Null hypothesis states null both cumulative distributions are similar. Rejecting the null hypothesis means cumulative distributions are different.In data science, it compares the cumulative distribution of events and non-events and KS is where there is a maximum difference between the two distributions. In simple words, it helps us to understand how well our predictive model is able to discriminate between events and non-events.
Suppose you are building a propensity model in which objective is to identify prospects who are likely to buy a particular product. In this case, dependent (target) variable is in binary form which has only two outcomes : 0 (Non-event) or 1 (Event). "Event" means people who purchased the product. "Non-event" refers to people who didn't buy the product. KS Statistics measures whether model is able to distinguish between prospects and non-prospects.