Data-Stream-Plugin (formerly Concept-Drift-Plugin)
Hi there
I ve downloaded the rapidminer-datastream-4.0beta plugin and put it into the plugins folder. But it seems there are no new operators available in RapidMiner. Are the corresponding operators already part of RapidMiner 5? If so, what are the names of the available operators for handling concept drift? Are there any sample processes for it?
Thank you and kind regards
Hagen
I ve downloaded the rapidminer-datastream-4.0beta plugin and put it into the plugins folder. But it seems there are no new operators available in RapidMiner. Are the corresponding operators already part of RapidMiner 5? If so, what are the names of the available operators for handling concept drift? Are there any sample processes for it?
Thank you and kind regards
Hagen
Find more posts tagged with
Sort by:
1 - 7 of
71
Hi haddock
Thanks for your help. I have already discovered that one. But this operator uses a fixed window, which makes no sense in presence of a stable concepts. Therefore it would be better to have an adaptive window. I just read about a method which is called drift detection method and is used to determine the point in a time-series at which a concept drift has occurred. Basically it monitors the classification error which will decrease over time in the absence of concept drift. The detection has two steps. First a warning(1) and then a point where it is almost certain(2) that a concept drift has occurred. If this happens the data instances between (1) and (2) are used for building a new prediction model. Sounds pretty easy, that is why I thought there might be an implementation for RapidMiner?
Best regards
Hagen
Thanks for your help. I have already discovered that one. But this operator uses a fixed window, which makes no sense in presence of a stable concepts. Therefore it would be better to have an adaptive window. I just read about a method which is called drift detection method and is used to determine the point in a time-series at which a concept drift has occurred. Basically it monitors the classification error which will decrease over time in the absence of concept drift. The detection has two steps. First a warning(1) and then a point where it is almost certain(2) that a concept drift has occurred. If this happens the data instances between (1) and (2) are used for building a new prediction model. Sounds pretty easy, that is why I thought there might be an implementation for RapidMiner?
Best regards
Hagen
Have you seen the series plugin for Rapidminer 5? Rapidminer 4.0 is only dimly remembered, and even less supported!