image

🎉 🎉 RAPIDMINER 9.10 IS OUT!!! 🎉🎉

Download the latest version helping analytics teams accelerate time-to-value for streaming and IIOT use cases.

CLICK HERE TO DOWNLOAD

Comma values - best predictive model suggestion

BamboBambo Member Posts: 6 Contributor I
edited December 2019 in Help

Hello, I would like to know what is the best predictive operator to solve my problem.

 

input_output_comma.PNG

 

So my main goal is to find the best predictive operator which could handle comma values (Input 1, Input 2, Input 3) and values in "Input 4" column (YES/NO) and predict values in "Output" column. I already tried Optimize Parameters(Evolutionary) with operators like [SVM kernel types: (radial, anova, epachenikov etc), Random Forest, Neural Net, Decision Tree] to find optimal values, but the average accuracy of prediction was low.

 

I would really like to know if someone could suggest some predictive operator which could produce the best accuracy for this kind of problem.

NOTE: The values in columns "Input 1, Input 2, Input 3" after adding up always equals 100% in every row if this helps.

Thanks very much in advance.

Tagged:

Answers

  • Pavithra_RaoPavithra_Rao Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 123  RM Data Scientist

    Hi @Bambo,

     

    You could try new 'Auto Model' feature to find the best predictive model for given data set.

     

    RapidMiner Auto Model

     

    Cheers,

    sgenzer
  • kypexinkypexin Moderator, RapidMiner Certified Analyst, Member Posts: 290   Unicorn

    Hi @Bambo

     

    For me, it would matter if the output column represents an arbitrary integer value or a categorical value (and in the latter case, how many unique categories you can have). If it's categorical, you could possibly use any non-linear algorithm like decision tree / RF / GBT. If it's numerical, then regression models would fit, but most likely you'll have to transform boolean column then.

     

    Generally speaking, you still have to experiment with different models and compare results. 

    sgenzeryyhuang
Sign In or Register to comment.