How to Deploy a Decision Tree on RapidMiner Server

Chiko

Hi,

I have created a decision tree process, trained and tested it in RapidMiner studio and saved the process on a process subfolder on my RapidMiner server. The next thing is I would like to deploy/productionize it on the server and give it an input file(excel) with unseen examples so that it can predicting. I do not know how to do this on RapidMiner Server.  Could someone assist please?

 

Thanks,

 

Chiko

Telcontar120

Hi there @Chiko!

 

This should be straightforward, assuming your new records to be scored have all the same data as your original training/testing dataset.  Probably the easiest way to accomplish it is as follows:

1. You'll need to use the "store" operator to store the model that you created in your first project on the server repository.
2. Then, you'll also need to store the dataset of records to be scored on the server repository as well.
3. Once they are both available on the server, you will write a process using the "retrieve" operator to first retrieve the dataset, (and make sure you do any data ETL or preprocessing), and then retrieve the model.
4. Finally you will use "apply model" operator to score the new dataset.  That will output the unscored records with the prediction and the confidence from your original model.

 

All of these processes will be written in Studio and then stored and executed on your Server.  

 

 

Telcontar120

Hi there @Chiko!

 

This should be straightforward, assuming your new records to be scored have all the same data as your original training/testing dataset.  Probably the easiest way to accomplish it is as follows:

1. You'll need to use the "store" operator to store the model that you created in your first project on the server repository.
2. Then, you'll also need to store the dataset of records to be scored on the server repository as well.
3. Once they are both available on the server, you will write a process using the "retrieve" operator to first retrieve the dataset, (and make sure you do any data ETL or preprocessing), and then retrieve the model.
4. Finally you will use "apply model" operator to score the new dataset.  That will output the unscored records with the prediction and the confidence from your original model.

 

All of these processes will be written in Studio and then stored and executed on your Server.  

 

 

Chiko

Thanks Telcontar120, works perfectly.