Time Series Forecasting for Data

glybenttaglybentta Member Posts: 6 Contributor I
edited November 2018 in Help

I have a marine dataset with 3 attributes (Year, Species [categorical], Catch). I need to forecast the catch of each individual species(there are 8 species) for next 5 years. Is it possible with Windowing operator? Can someone please help me out with this? Thanks in advance

Answers

  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn

    Of course it is possible :-), with RapidMiner it almost always is.  

    However, in these types of cases, you may find that you need to create 5 separate forecast models.  Of course, if the data structure is the same for each, then you can easily reuse the same modeling setup and just loop through the setup by selecting only examples for each species separately for each run.

    If you install the Time Series extension, there should be a new template available on the Studio startup page (or you can navigate there from your repository under Samples under Templates) which will help you get your first process set up.  Take a look at the attached screenshot.

    time series tutorial.PNG

     

     

    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
  • Thomas_OttThomas_Ott RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,761 Unicorn
  • glybenttaglybentta Member Posts: 6 Contributor I

    Thank you Brian T for you quick reply. I went through the above mentioned templete and the link provided by Thomas Ott. I understood the concept of windowing, but I could only forecast the values which are already present in my ExampleSet. I have data from 1980-2016. So with the windowing operator with parameters as window size = 5, step size = 1, horizon= 1, I could only get the forecasted value of 2016 which is already present in my dataset. I tried changing the forecast models(Linear regression, SVM etc) but the forecasted output dint change at all. I dont know wether I had gone wrong somewhere. What if I wanted to forecast value of 2017 using the previous 5 years data?

  • Thomas_OttThomas_Ott RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,761 Unicorn

    Take a look at the sample processes I posts here: http://community.rapidminer.com/t5/RapidMiner-Studio/Recall-Error/m-p/37302#U37302

     

    This process was taking out of Vijay and Bala's book (Chp 10) and uses the existing time series to forecast into the future based on the model it created. It's really neat but you'll have to tweak it to your data set. 

Sign In or Register to comment.