RapidMiner slow to launch

raiyaniraiyani Member Posts: 3 Contributor I
edited November 2018 in Help
Hello,

I am new to RapidMiner and am facing the following issue

Platform: Java (Slackware Linux)
Version: 5.3.000
Memory: 12 GB
Processor cores: 8

I was using RapidMiner without a hitch. Now, when I try to launch it with an existing process, I am only offered a blank sheet (new process). Also, if I drag any operator to the canvas, the operator is greyed out and the Help information box shows an hourglass.

I am using RapidMiner for text mining, and I have been encountering this issue since I started using the following process:
  • Read an Excel file with ~3,000 rows
  • Process the text in the Excel file  (tokenization, change case, stemming, etc.)
  • Clustering using k-NN
Thanks,
Samir

Answers

  • Marco_BoeckMarco_Boeck Team Lead Software Engineering Moderator, Employee, Member, University Professor Posts: 1,780   RM Engineering
    Hi,

    please update to RapidMiner version 5.3.005. Unfortunately, a bug slipped into the 5.3.000 release which can cause the symptons you describe.

    Regards,
    Marco
  • raiyaniraiyani Member Posts: 3 Contributor I
    Thank you, Marco. I updated RapidMiner to version 5.3.005 as you've suggested.

    Here is what happens now:
    • When I launch, RapidMiner, I can select from the list of recent processes
    • If I select an existing process, it takes 2-3 minutes to load the process; after that, it works fine
    • If I start a new process, I can add new operators (nothing is greyed out); but when I run the process, it takes 2-3 minutes for the results to show; after that, it works fine/li]
    Thanks for your help.
  • Nils_WoehlerNils_Woehler Employee, Member Posts: 463  Guru
    Hi there,

    this is probably a problem of an extension you have installed.
    Please go to your .RapidMiner5 folder located in your system user folder (e.g. /home/user/.RapidMiner5) and remove all files that start with vlperspective.
    This will reset your perspectives to the default settings but should improve the time needed to switch to the result perspective.

    Best,
    Nils
  • raiyaniraiyani Member Posts: 3 Contributor I
    Hi Nils,

    Removing the vlperspective* files in the .RapidMiner5 folder does not help at all.

    Thanks,
    Samir
  • MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869   Unicorn
    Hi Samir,

    are you using a 64bit Java version? Otherwise, RapidMiner can't use more than 2GB of RAM. Talking about RAM, can you please check how much RAM RapidMiner is allowed to use? You can check that e.g. by opening the System Monitor View in RapidMiner and have a look at the Max value. If it is too low, RapidMiner can suffer from the slowdowns that you describe. You can configure the max RAM in too ways:
    - if you are using the RapidMinerGUI.sh start script from the scripts folder, edit it and adjust (and uncomment) the MAX_JAVA_MEMORY variable (value in MB).
    - if using the script, you can also set that variable as a shell variable
    - if you are starting RapidMiner via java -jar, add the -Xmx parameter, e.g. java -Xmx4096m -jar rapidminer.jar

    If memory is not an issue, let's talk about the data: how much data do you load, and which kind of data do you display in the end? If it is text data, it can be arbitrarily slow. In that case, please try to remove the text attributes (and I really mean attributes of type text, not nominal) prior to displaying the data.
    Furthermore, RapidMiner calculates some statistics on the data. Usually this is not done for large datasets (in terms of examples), but if you have very many attributes, as typical for text mining, it may take some time. You can define up too how many examples the statistics are calculated in the RapidMiner settings in the preferences. The option is located in the Gui tab as rapidminer.gui.max_statistics_rows

    By saying that it takes 2-3 minutes to display the data, I suppose you mean that the process finishes, and afterwards it needs some time to show the Results perspective? Just want to make sure that we are talking about the same things, but you are certainly aware that actually *running* the process can potentially take a veeery long time. But we are talking just about the loading of the results, right?

    Concerning the process loading, please right-click your repository and select "Open in File Browser". Do you see any process files which are significantly larger than 10kb? That should not be the case, unless you have built *very* complex processes.

    Good luck!

    ~Marius
Sign In or Register to comment.