Looking for some help on multi-instance and time series forecasting
Hi All,
I am a student that is relatively new to WEKA but have steadily been teaching myself with textbooks/lessons. I am working on a project that *I think* requires the combination of both multi-instance and time series forecasting in one model. This has proved to be quite difficult thus far.
I am working with a data set consisting of ~70 companies and ~70 attributes. This is then repeated each year for a period of 15 years. I also have a fair bit of missing data. My project requires me to compare a few classifiers based on their predictive ability on a binary (1,0) class.
Is such a thing possible? I note that when I "bag" the data for use with multi-instance, the forecasting tab is no longer usable (date attribute is in the bag so I'm guessing that's why). Any idea on how to achieve this or any pointers would be very helpful. Thank you in advance.
My .arff file looks something like this:
@relation companyfinancials
@attribute Company {Company1,Company2,...,Company70}
@attribute Year date yyyy
@attribute profit numeric
@attribute cashflow numeric
...
@attribute class {1,0}
@data
Company1,2015,380.07,142.08,...,1
Company2,2015,2192.74,439,...,0
...
Company70,2015,3545.61,10,...,1
Company1,2014,380.07,142.08,...,1
Company2,2014,2192.74,439,...,0
...
Company70,2014,3545.61,10,...,1
...
Company1,2001,380.07,142.08,...,1
Company2,2001,2192.74,439,...,0
...
Company70,2001,3545.61,10,...,1
