I forgot to mention: you wouldn’t be able to do this automatically in something like the
Forecasting panel of the Explorer. You’d have to create the data for each model using some
other method and then build the models in the Classify panel. You’d probably have to write
a script to generate the data to train the models.
On 13/09/2017, at 1:39 PM, Eibe Frank
By “correspond” I didn’t mean that they need to be the same. Is record N for patient X
from the same time period (e.g., year) as record N for patient Y? This is the case for the
balanced panel example here: https://en.wikipedia.org/wiki/Panel_data
In that case, the suggestion I had may apply: you could try building a classification
model from all year N observations, another model from all year N and N+1 observations
(where the data for a patient from N and N+1 is used to make a *single* row/instance for
WEKA), and a third model from all year N, N+2, and N+3 observations (where again the data
for a patient from N, N+1, and N+2 is used to make a *single* row/instance for WEKA). This
would not give you models that are specific to any individual, but you could use them to
obtain classifications for all individuals and years. The main thing is to avoid using the
future to predict the past.
Note that I have not worked with panel data before so please take the above with a grain
On 13/09/2017, at 12:58 PM, alkunany
The number of records is same for each patient , and the record N of patient
X is not correspond to record N of patient Y ( its panel data) each patient
has his/her own records.
Assume I will include only test1 , WEKA’s timseriesforecasting works only
on records for one patient at time , however each patient has own different
records ? and I want to predict the case of each patient based on his
records ( panel data).
Could you have any suggestion please ?
Please I am waiting your positive response .
Sent from: http://weka.8497.n7.nabble.com/
Wekalist mailing list
Send posts to: Wekalist(a)list.waikato.ac.nz
List info and subscription status: https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html