Open siddij3 opened 6 months ago
Hi @siddij3 (sorry for the super late response! )
I do not know where you got this specific list from (maybe you compiled it yourself?), but for usage in the kind_to_fc_parameters
, this list has the wrong format. A your basically have a single dictionary for all the number_of_maxima
settings, python will go ahead and combine them all into one key (you can go ahead and look at the resulting settings e.g. with a print
statement and you will see, that even before you pass it to tsfresh, there is only a single entry (the last one) in it).
The format expected by tsfresh is a list of dictionaries. So for your use case:
settings = {
"mean_n_absolute_max": [
{
"number_of_maxima": 7,
},
{
"number_of_maxima": 9,
},
{
"number_of_maxima": 11,
},
{
"number_of_maxima": 13,
},
{
"number_of_maxima": 15,
},
{
"number_of_maxima": 17,
},
{
"number_of_maxima": 19,
},
{
"number_of_maxima": 21,
},
{
"number_of_maxima": 23,
}
],
}
With this, it will work as expected.
Thanks for replying! I compiled the list myself after filtering through the features based on their results and the physical implications of the features. I completely overlooked the dictionary key overlapping. Thanks for pointing that out!
I ran kind_to_fc_parameters to extract custom parameters from the extract_features() function with a set of parameters, and in them was the following:
When looking at the returned columns, it only returned number_of_maxima_23 for all the timeseries columns.
Intial timeseries columns:
Extracted:
Environment: