Kmeans algorithm test for Apache mahout

Mahout is a data mining package in Hadoop, although it is now generally used with spark mlib, but in order to make comparisons, think of the mahout algorithm to verify the test.

Mahout installation is very simple, just need to decompress and then the following configuration can be.

# Mahout
Export mahout_home=/home/ndscbigdata/soft/mahout-0.11.0
Export path= $MAHOUT _home/bin: $PATH
Export classpath= $CLASSPATH: $MAHOUT _home/lib

Because the algorithm has a lot of blogs or books are introduced, so it is not repeated here.

Among them, the Kmeans algorithm test adopts the popular "Syntheticcontrol" data set of the Internet to realize. The mahout provides three algorithms for this data set, as follows:

The result of the Kmeans test is as shown in the figure.

The results are analyzed and compared, and it is found that the efficiency of data mining based on Hadoop is very low, and it will store the intermediate clustering results in disk every time.