Data Science musings from today: the validity of your ability to train models on your own data is often dependent on what actions you make available to your users.
-
Show this thread
-
If you force users into having to come up with hacks to use your product/service the way that feels natural to them, you will train your models based on hacks. You can't derive insights that way, you can only build more convoluted hacks.
1 reply 1 retweet 11 likesShow this thread -
User behavior is often deterministic, not stochastic. If you can describe easily your desired outcome in words, applying statistical techniques should not be where you're starting.
1 reply 1 retweet 7 likesShow this thread -
Here's an example. Suppose you're making, I don't know, a hotel booking service, and you allow your users to enter a price range. How hard is it for users to manipulate that price range to meet their needs? Is it annoying? Natural? Does this reflect what people actually want?
1 reply 0 retweets 2 likesShow this thread -
If you have a bad user flow, you have to take this into account. This is kind of an equivalent of right-censored data: you're artificially cutting off behavior based on what you make available.
1 reply 0 retweets 5 likesShow this thread
So what you actually need to do is precondition your training data with survivorship models first.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.