Questions about Snowflake asked by my Students on Udemy (Q1-Q5)

These are all real-life questions I’ve been asked by students in the Q&A sections of my Udemy courses about Snowflake. I’ll just keep the anonymity of the sender and I’ll only reformat them a bit, to make them more readable. I will publish five questions per post, stay connected!

[Q1] In a Forecasting ML model in Snowflake Cortex, how come that the Lower Bound is negative if all my values are Positive?

I suspect you refer to the prediction or confidence intervals…

It may happen, because these are statistical measurements and you have to consider a margin of error.

Let’s say you trained your model with all values between 1 and 100. The algorithm (and not just in Snowflake) could suggest a prediction/confidence interval of let’s say -10 .. 120.

It tells you there is a probability that you may get in real life a few values below 0 as well, in this case.

And something else: these algorithms deal with numbers. They have no idea if you deal with some metric that can never go below zero. If that’s the case, you could simply disregard whatever negative values or intervals.