"windowing a discontinuous data set"

pop

Hello,

I have a set of intraday data over several days. I am using the windowing operator on it.
The window size is small with regard to the number of intraday data so most of the windowed examples are made of data from the same day, but this operator will, as expected, create some examples that include data from 2 different days.
I would like to remove those examples as they make no sense.
Any idea for creating this kind of filter??

Thank you very much for your help.

haddock

Hi there Pop,

You can probably get what you want by using the "Date to Numerical" operator, which can strip the components from adate_timeattribute. In the following example data is filtered by the minute of the timestamp, leaving only those from a quarter past the hour.

<宏/ >




< parameter key="repository_entry" value="Past_Deltas"/>


< parameter key="attribute_name" value="DateTime"/>
< parameter key="date_type" value="date_time"/>
< parameter key="date_format" value="yyyy-MM-dd hh:mm:ss.0"/>


< parameter key="attribute_name" value="DateTime"/>
< parameter key="time_unit" value="minute"/>
< parameter key="keep_old_attribute" value="true"/>


< parameter key="condition_class" value="attribute_value_filter"/>
< parameter key="parameter_string" value="DateTime_minute = 15"/>

Have fun 8)

pop

Hello Haddock,
Thank you very much for your reply.
I was thinking about something like this:
before windowing I have one attribute called date.
After windowing I have attributes date-99, date-98,...,date-0.
I would basically need to create a filter that keep the example if all values of date-99,date-98,....,date-0 are all the same and remove it otherwise.
Does it make any sense?
thank you

haddock

Hi Again,

You could use that operator to strip out the day from just the first and last timestamps, and trash examples where they weren't the same. Worth a try anyway..

Good luck !