Skip to content

mikusmi/clueminer

 
 

Repository files navigation

Clueminer

Build Status

Clueminer is a platform for interactive data-mining with special focus on clustering algorithms.

Building latest development version

Prerequisites:

  • git

  • Java 8 or newer

  • Maven

  • on Debian/Ubuntu Linux:

       apt install maven libjhdf5 git
       git clone git://github.com/deric/clueminer.git
       cd clueminer
       git submodule init
       git submodule update
       mvn clean install
    
  • Once build you can run Clueminer via Maven

cd modules/application
mvn nbm:cluster-app nbm:run-platform
  • or use generated bin file
bash modules/application/target/clueminer/bin/clueminer

From NetBeans

  1. open the clueminer directory (which is a maven module) -- this module is called clueminer-parent
  2. from dependent modules of clueminer-parent open clueminer-app
  3. execute "Build with dependencies"
  4. now you can run the main application (run module clueminer-app)

How to increase heap size

In application/src/main/resources/clueminer.conf adjust Java options:

default_options="--branding clueminer -J-Xms24m -J-Xmx2048m"

Note: increasing heap size J-Xmx to values bigger than is your actual physical RAM will cause serious preformance issues!

On Unix systems you can find out your memory size with this command:

echo $(( $(awk '/MemTotal/{print $2}' /proc/meminfo) >> 10 ))m

Forking

  1. click on fork button on github
  2. add upstream repository
 $ git remote add upstream https://github.com/deric/clueminer.git
  1. from time to time merge with upstream
$ git fetch upstream
$ git checkout master
$ git merge upstream/master

Benchmarks

Benchmarks of clustering algorithms are located in modules/clustering-benchmark

In order to run benchmarks build an asssembly with Maven:

$ cd modules/clustering-benchmark
$ mvn assembly:assembly

and run benchmarks (might be computationally expensive):

$ java -jar target/*-jar-with-dependencies.jar

R support

In order to enable R code execution JRI is needed.

Debian/Ubuntu:

apt install r-cran-rjava

HDF support

Library libjhdf5 is needed to be present on library path.

Debian/Ubuntu:

apt install libjhdf5-jni

OpenGL support

For OpenGL visualizations you'll need native extensions

Debian/Ubuntu

jzy3d is using jogl library which has native bindings:

sudo apt-get install libjogl-java

R support

Currently there are two possibilities how to execute R code from Java:

  • TCP connection to RServe
  • load dynamic library libjri.so
    • for Debian: apt install r-cran-rjava, make sure R_HOME is set
    • update classpath export CLASSPATH=.:/usr/lib/R/site-library/rJava/jri/

About

interactive clustering platform

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Java 100.0%