Hello all, I will like to know if anyone has any e...
# generalg
Galvan Goh
04/11/2024, 2:06 AMHello all, I will like to know if anyone has any experience working with multiple time series, whereby you pass each series through some tests and the result of the tests provides an indication that the series belongs to a certain group (something like time series clustering but without training a model). I have a panel data set containing a lot of series which are different in time periods and scale. I want to separate all of them into different groups and apply the appropriate preprocessing steps to them before performing forecasting. TIA!
m
Mariana Menchero
04/11/2024, 2:30 AMHi @Galvan Goh we currently don't have any functionality specifically designed for time series classification.
Mariana Menchero
04/11/2024, 2:32 AMI would recommend doing a time series decomposition and then classify the series based on whether they have seasonality, increasing trend, decreasing trend, etc.
g
Galvan Goh
04/11/2024, 2:45 AMHello @Mariana Menchero thank you for the suggestion. I will try that out.
Galvan Goh
04/30/2024, 6:07 AM@Mariana Menchero I have multiplicative time series which are all negative and also mix of positive and negative values. Decomposition will require time series to be additive. The solution will be to apply log or box-cox transformation but the presence of negative values in the data will cause the function to throw errors. Can you recommend what can I do in this situation?
m
Mariana Menchero
05/01/2024, 5:15 AMHi @Galvan Goh if you need to apply a log or a Box-Cox transformation but your series are all negative or have negative values, maybe you add a constant first to the data so that all values are positive and then do the transformation. Keep in mind that if you do this, at the end you need to subtract the constant that you just added.
Mariana Menchero
05/01/2024, 5:15 AMas a constant, you could add the absolute value of the minimum plus 1