Topic
IC4NOTICE: developerWorks Community will be offline May 29-30, 2015 while we upgrade to the latest version of IBM Connections. For more information, read our upgrade FAQ.
8 replies Latest Post - ‏2013-09-12T12:48:00Z by mato4
mato4
mato4
8 Posts
ACCEPTED ANSWER

Pinned topic SPSS Analytics Toolkit installation and sample applications

‏2013-08-22T11:48:36Z |
Hi all, 
I would need some advice with installation of SPSS Analytics Toolkit. I have virtual server with InfoSphere Streams and virtual server with SPSS(windows). 
  1. Is it possible to have 2 servers - one with Streams and one with SPSS and use SPSS Analytics Toolkit? Or do i need to install everything on one server?
  2. I know that SPSS Analytics Toolkit install file is in SPSS for Linux. Can I use SPSS on windows or this toolkit can only work with SPSS on Linux? 
I have .pdf document IBM SPSS Analytics Toolkit for InfoSphere Streams Version  1.0 but I am little confused about installation procedure. If somebody installs this toolkit and can help me I would really appreciate that. Thanks!

Matej

  • Kevin_Foster
    Kevin_Foster
    98 Posts
    ACCEPTED ANSWER

    Re: SPSS Analytics Toolkit installation

    ‏2013-08-22T19:58:22Z  in response to mato4

    I went to an expert on this topic:

     

    There is a product named SPSS Modeler Server and also a product named SPSS Modeler Solution Publisher.

    You need to install the second one on the same Linux machine as your Streams installation. That will give you the SPSS run-time that you need for the SPSS operator to function inside of Streams.

     

    You can read more here:

    http://pic.dhe.ibm.com/infocenter/streams/v3r1/topic/com.ibm.swg.im.infosphere.streams.mining-toolkit.doc/doc/overview.html?resultof=%22%73%70%73%73%22%20

     

    -Kevin

     

     

     

  • JDThorson
    JDThorson
    3 Posts
    ACCEPTED ANSWER

    Re: SPSS Analytics Toolkit installation

    ‏2013-08-22T20:01:34Z  in response to mato4

    My guess is that you are assuming the engine behind the toolkit is SPSS Modeler Server, it is not.

    The SPSS Analytics Toolkit uses SPSS Modeler Solution Publisher 'in process' from the SPSSScoring operator, there is not a remote SPSS server involved here.  The various libraries used to implement the process and model nodes in your SPSS Modeler scoring branch are loaded dynamically so there is a requirement to have the SPSS Modeler Solution Publisher install directory visible to all nodes running an instance of the scoring operator from this toolkit.

    (A) the toolkit is delivered in the Modeler Solution Publisher, install the RedHat version of this product into your envrionment

    (B) the document describes use of the toolkit which is a fairly standard Streams toolkit tarball

    Please feel free to contact me if you have additional questions.

    • mato4
      mato4
      8 Posts
      ACCEPTED ANSWER

      Re: SPSS Analytics Toolkit installation

      ‏2013-08-26T08:44:20Z  in response to JDThorson

      Hi John,

      thank you for reply. So I will install Solution Publisher on server with Streams. Where can I create my SPSS model? Can I create model on another server or do I need to do everything on the same server?

      Matej

      • JDThorson
        JDThorson
        3 Posts
        ACCEPTED ANSWER

        Re: SPSS Analytics Toolkit installation

        ‏2013-08-26T13:06:50Z  in response to mato4

        SPSS Modeler Client is where one designs the plan for training, evaluating and scoring with the predictive models your application requires. 

        The SPSS Modeler Client product uses data mining techniques to train the predictive model.  This is usually a process that involves a significant amount of historic data.  Some percentage of this historic data is usually 'held back' for model evaluation.  This is not a one-time 'create' and you will want to keep your models current and accurate by a planned model refresh schedule.  Depending on your hardware requirements and the duration of this process one may choose to do the training and evaluation all in SPSS Modeler Client or use SPSS Modeler Server.  So the answer here is the hardware applied will be in support of the Streams application but not part of the Streams instance.

        The majority of Streams applications and streaming data requirements have little use for a long-running process like the predictive model training and evaluation described above.  For these reasons the SPSS toolkit for InfoSphere Streams is focused on scoring with the trained predictive models and although it has mechanisms to facilitate an automated 'refresh' of the models in use it does not include model build and evaluation support.

        • mato4
          mato4
          8 Posts
          ACCEPTED ANSWER

          Re: SPSS Analytics Toolkit installation

          ‏2013-08-30T14:41:38Z  in response to JDThorson

          Hi,

          I installed SPSS Analytics Toolkit, then I tried to launch sample scoring application in InfoSphere Streams Studio and I got this error:

          ---- Indexing toolkit for project SPSS started ---- August 30, 2013 10:36:54 AM EDT
          ---- Indexing toolkit for project SPSS completed in 1.298 seconds ----

          ---- SPL Build for project SPSS started ---- August 30, 2013 10:36:56 AM EDT

          Building main composite: SPSSScoringExample using build configuration: Distributed

          /opt/ibm/InfoSphereStreams/bin/sc -M SPSSScoringExample --output-directory=output/SPSSScoringExample/Distributed --data-directory=data -t /opt/ibm/com.ibm.spss.streams.analytics --no-toolkit-indexing --no-mixed-mode-preprocessing

          Checking constraints...
          Creating types...
          Creating functions...
          Creating operators...
          CDISP9007E ERROR: Cannot open file '/opt/ibm/com.ibm.spss.streams.analytics/com.ibm.spss.streams.analytics/SPSSScoring/SPSSScoring_h.pm' for writing.
          CDISP0510E ERROR: Code generation has failed for code generation template '/opt/ibm/com.ibm.spss.streams.analytics/com.ibm.spss.streams.analytics/SPSSScoring/SPSSScoring_h.cgt'.

          ---- SPL Build for project SPSS completed in 2.899 seconds ----
           

          Can you please help me with this problem? Thanks!

          • JDThorson
            JDThorson
            3 Posts
            ACCEPTED ANSWER

            Re: SPSS Analytics Toolkit installation

            ‏2013-08-30T15:09:56Z  in response to mato4

            Your procedures must make sure the installed files and directories are 'executable' and any file we are to write to (published content in this case) are 'writable' for use by the toolkit.

            • This reply was deleted by mato4 2013-09-12T12:48:48Z.
              • This reply was deleted by mato4 2013-09-12T12:48:09Z.
  • mato4
    mato4
    8 Posts
    ACCEPTED ANSWER

    Re: SPSS Analytics Toolkit installation

    ‏2013-09-03T10:56:26Z  in response to mato4

    Hi,
    can you help me to configure SPSSRepository operator? I have simple application only with this operator and I configured parameters. But .str filed is not downloaded. I checked logs but there is nothing. I suppose that parameters are correct because when I intentionally typed wrong password or host address or port then I saw error in logs. Maybe there is somewhere problem with SPSS Collaboration and Deployment server but I can see in the repository my .str file and also RSS feeds are published when I update this file. Thanks!

    Matej

     

    spl code:

    stream<rstring filePath> strFile = com.ibm.spss.streams.analytics::SPSSRepository(){
                param
                    cdsServer: "http://192.168.128.130:9080";
                    userID:"admin";
                    password:"psw";
                    resourceURI:"spsscr:///?id=7f0000010087ca5c00000140c977f2d99d67";
                    targetFilePath:"/home/streamsadmin/Desktop/workspace/SPSSRepositoryPublishScoring/data";
                    versionLabelName:"LATEST";
                    detectionPeriod: 10;
            }

    SOLVED:

    You just need to have running stream application and then save spss stream project to repository. If you start stream application it won´t download spss stream from repository until you save it from spss modeler to repository.

  • mato4
    mato4
    8 Posts
    ACCEPTED ANSWER

    Re: SPSS Analytics Toolkit installation and sample applications

    ‏2013-09-12T12:48:00Z  in response to mato4

    Hi,
    I have still problem with scoring operator. Repository and publish operators work fine. I am working with sample project, I´ve changed only parameters. This error occures only for scoring operator. Thanks for help.

     

    ---- SPL Build for project spss started ---- September 1, 2013 1:25:25 PM EDT

    Building main composite: SPSSScoringExample using build configuration: Distributed

    /opt/ibm/InfoSphereStreams/bin/sc -M SPSSScoringExample --output-directory=output/SPSSScoringExample/Distributed --data-directory=data -t /opt/ibm/com.ibm.spss.streams.analytics --no-toolkit-indexing --no-mixed-mode-preprocessing

    Checking constraints...
    Creating types...
    Creating functions...
    Creating operators...
    Creating PEs...
    Creating application model...
    Building binaries...
     [CXX-type] tuple<rstring s_sex,int64 baseSalary,int64 bonusSalary>
     [CXX-type] tuple<rstring s_sex,int64 baseSalary,int64 bonusSalary,in...,float64 confidence>
     [CXX-type] tuple<rstring fileName>
     [CXX-operator] data
     [CXX-operator] notifier
     [CXX-operator] scorer
    src/operator/scorer.cpp: In constructor 'SPL::_Operator::scorer$OP::scorer$OP()':
    src/operator/scorer.cpp:43: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp: In member function 'virtual void SPL::_Operator::scorer$OP::process(SPL::Tuple&, uint32_t)':
    src/operator/scorer.cpp:223: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp: In member function 'virtual void SPL::_Operator::scorer$OP::process(const SPL::Punctuation&, uint32_t)':
    src/operator/scorer.cpp:257: error: 'TRUE' was not declared in this scope
    src/operator/scorer.cpp: In member function 'bool SPL::_Operator::scorer$OP::field_proc(int, int*)':
    src/operator/scorer.cpp:374: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:377: error: 'TRUE' was not declared in this scope
    src/operator/scorer.cpp: In member function 'bool SPL::_Operator::scorer$OP::prepareModel()':
    src/operator/scorer.cpp:442: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:449: warning: deprecated conversion from string constant to 'char*'
    src/operator/scorer.cpp:455: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:460: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:463: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:473: warning: deprecated conversion from string constant to 'char*'
    src/operator/scorer.cpp:479: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:484: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:487: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:499: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:504: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:508: error: 'FALSE' was not declared in this scope
    src/operator/scorer.cpp:510: error: 'TRUE' was not declared in this scope
    src/operator/scorer.cpp: In member function 'bool SPL::_Operator::scorer$OP::validNewImageOp(int, const char*)':
    src/operator/scorer.cpp:521: error: 'TRUE' was not declared in this scope
    src/operator/scorer.cpp:528: error: 'FALSE' was not declared in this scope
    make: *** [build/operator/scorer.o] Error 1
    CDISP0141E ERROR: Compilation of the generated code has failed.

    ---- SPL Build for project spss completed in 19.32 seconds ----