thanks for posting on time series... I've been looking for good tutorials for the past two months, and your videos are actually what I was looking for THANK YOU !
Hey, quick question, not sure if you will mention it in the following videos of the series. In the previous videos you mentioned techniques to find whether data is stationary or not (by removing trend, seasonality, and even transforming the data with boxcox, log functions, etc). Now we are decomposing the "original" time series, but shouldn't we transform data first? I mean, why should we run statistical tests for stationary data if the decomposition process doesn't use that? Is it related to assumptions? Can we only assume the time series decomposition models work if data is stationary? And thanks for the content, mate! Really awesome and didatic
Hey, I understand the confusion! Decomposition is more of an analysis tool for the time series. You can transform the data first and then do it to determine if the data is stationary through decomp. Hope that makes sense?
you could transform the multiplicative seasonality into additive seasonality using log transform, then do decomposition. Or vice versa using differencing!
Glad to see the lighting is better than the first video but it's still too bright at the window's side. Found you on medium, looking forward to seeing more.
Sorry, I cant offer advice as I have no experience in this area. My de-facto would be to apply the same way as everybody else, on google, linkedin etc.
thanks for posting on time series... I've been looking for good tutorials for the past two months, and your videos are actually what I was looking for
THANK YOU !
thanks for the kind words!
Please make a machine learning crash course too! This was the exact thing I was looking over all across TH-cam
Hey, I will have a think! ML is so broad, so would take awhile to cover everything!
Hey, quick question, not sure if you will mention it in the following videos of the series. In the previous videos you mentioned techniques to find whether data is stationary or not (by removing trend, seasonality, and even transforming the data with boxcox, log functions, etc). Now we are decomposing the "original" time series, but shouldn't we transform data first?
I mean, why should we run statistical tests for stationary data if the decomposition process doesn't use that? Is it related to assumptions? Can we only assume the time series decomposition models work if data is stationary?
And thanks for the content, mate! Really awesome and didatic
Hey, I understand the confusion! Decomposition is more of an analysis tool for the time series. You can transform the data first and then do it to determine if the data is stationary through decomp. Hope that makes sense?
Nice. What if you have additive trend and multiplicative seasonality or vice versa, how would we decompose that? it would be tricky.
you could transform the multiplicative seasonality into additive seasonality using log transform, then do decomposition. Or vice versa using differencing!
Glad to see the lighting is better than the first video but it's still too bright at the window's side.
Found you on medium, looking forward to seeing more.
Thanks for the feedback! Will definitely invest in improving the lighting.
@@egorhowell the lighting is perfect... dont mind this dude. great vid
@@virgenalosveinte5915 haha thanks! My recent videos a lot better :)
Thank you! Clear explanation and nice presentation
Glad it was helpful!
Well explained, thanks!
Glad it was helpful!
Great video keep it up
Thank you Soran!
How do I apply for Data scientist jobs ? pov i am from India and interested to work abroad
Sorry, I cant offer advice as I have no experience in this area. My de-facto would be to apply the same way as everybody else, on google, linkedin etc.