"[SOLVED] Problems with Set Role and ECODB in Java Code"

DGSunican

Hello all,

I am having a problem integrating RapidMiner in my Java Code. I am using 5.3.013 version.

What i want to do is the next process:
1. Read a Weak .arff file.
2. Set as label the last nominal attribute, named "note"
3. Apply the ECODB algorithm in order to detect class-outliers
4. Save the resulting set, with the instances marked as outliers, in a new .arff file

Following the documentation in the RapidMiner Wiki, i have implement the following Java method to perform the mentioned steps:

try{
            //initialized RapidMiner
            RapidMiner.init();
            
        } catch(Exception e){System.out.println("Exception initializating Rapidminer, method RapidMiner.init(): "+e);}
        
        /*
         * Creates a new process. This process will contain the
         * whole sequence of reading the .arff Weka file, making the
         * outlier instances and finally saving the .arff Weka file with
         * the marked instances
         */
        com.rapidminer.Process outlierProcess = new com.rapidminer.Process();
        
        try{
            //Operator for reading the .arff Weka file
            Operator initialDatasetECODB = 
                    OperatorService.createOperator(com.rapidminer.operator.io.ArffExampleSource.class);
            
            //Operator to perform EDOCB algorithm
            Operator ecodb =
                    OperatorService.createOperator(com.rapidminer.operator.preprocessing.outlier.EcodbOperator.class);
            
            //Operator for saving the markedDataset
            Operator markedDatasetECODB =
                    OperatorService.createOperator(com.rapidminer.operator.io.ArffExampleSetWriter.class);
            
            //Operator for labeling the initial dataset
            Operator labeling =
                    OperatorService.createOperator(
                    com.rapidminer.operator.preprocessing.filter.ChangeAttributeRole.class);
            
            
            //Configures the "read arff" operator to read the dataset file with ID assigned per instance
            initialDatasetECODB.setParameter(
                    com.rapidminer.operator.io.ArffExampleSource.PARAMETER_DATA_FILE, 
                    //this.directorio.getAbsolutePath()+"\\originalDataSet.arff");
                    "C:/originalDataSet.arff");

           
            //Configures the ECODB operator
            ecodb.setParameter(
                    com.rapidminer.operator.preprocessing.outlier.EcodbOperator.PARAMETER_NUMBER_OF_NEIGHBORS, 
                    ""+this.k+"");
            ecodb.setParameter(
                    com.rapidminer.operator.preprocessing.outlier.EcodbOperator.PARAMETER_NUMBER_OF_Class_OUTLIERS,
                    ""+this.numOutliers+"");
            
            //Configures the operator to save the marked dataset
            markedDatasetECODB.setParameter(com.rapidminer.operator.io.ArffExampleSetWriter.PARAMETER_EXAMPLE_SET_FILE,
                    this.directorio.getAbsolutePath()+"\\markedDataset.arff");
            
            //Configures the operator for labeling the initial dataset
            labeling.setParameter(
                    com.rapidminer.operator.preprocessing.filter.ChangeAttributeRole.PARAMETER_TARGET_ROLE,
                    Attributes.LABEL_NAME);
            labeling.setParameter(
                    com.rapidminer.operator.preprocessing.filter.ChangeAttributeRole.PARAMETER_NAME,
                    "note");
            
            //Adds the operators to the process
            outlierProcess.getRootOperator().getSubprocess(0).addOperator(initialDatasetECODB);
            outlierProcess.getRootOperator().getSubprocess(0).addOperator(ecodb);
            outlierProcess.getRootOperator().getSubprocess(0).addOperator(markedDatasetECODB);
            outlierProcess.getRootOperator().getSubprocess(0).addOperator(labeling);

            
            
            
            //Conects the reader operator to the ecodb operator
            try{
            initialDatasetECODB.getOutputPorts().getPortByName("output")
                    .connectTo(labeling.getInputPorts().getPortByName("example set input"));
            }catch(Exception e){System.out.println("Error in initialDataset operator");}
            try{
            labeling.getOutputPorts().getPortByName("example set output")
                    .connectTo(ecodb.getInputPorts().getPortByName("example set input"));
            
            }catch(Exception e){System.out.println("Error in labeling operator: "+e);}
            try{
            ecodb.getOutputPorts().getPortByName("example set output")
                    .connectTo(markedDatasetECODB.getInputPorts().getPortByName("input"));
            }catch(Exception e){System.out.println("Error in ecodb  operator");}
            
            //Run the whole process
            outlierProcess.run();
            
        }catch (Exception e){ System.out.println("Exception trying to execute the ECODB RapidMiner process "+e);}
        

When i execute it, i get the following Exception:

Exception trying to execute the ECODB RapidMiner process com.rapidminer.operator.UserError: No data was delivered at port Detect Outlier (COF).example set input (connected). 

It seems that the output port of the Set_Role class, "Example output set", do not return nothing. I have tried to connect, directly, the ECODB class to the readArff class, and they seems to connect correctly, but of course ECODB launch en exception requiering to stablish a label attribute, and it only can be done with the Set_Role operation, which do not work correctly.

I hope someone in this forum could help me with this issue.

Thank you very much.

Marco_Boeck

Hi,

Please have a look at the sticky thread in the development forum: http://rapid-i.com/rapidforum/index.php/topic,5807.0.html

What I always recommend is to not create your own process by hand (i.e. don't create operators yourself) but to rather use a predefined process which has been created via RM Studio GUI. That makes things A LOT less prone  to errors.

Regards,
Marco

DGSunican

Marco Boeck wrote:
Hi, Please have a look at the sticky thread in the development forum: http://rapid-i.com/rapidforum/index.php/topic,5807.0.html What I always recommend is to not create your own process by hand (i.e. don't create operators yourself) but to rather use a predefined process which has been created via RM Studio GUI. That makes things A LOT less prone  to errors. Regards, Marco

Thank you so much Marco, it worked for me. Problem solved!