KMeans

KMeans Clustering

A C++ program that performs KMeans clustering and returns the accuracy

Ensure the path for mlpack has been set:

By default mlpack is installed to /usr/local/, if you have installed somewhere specify that path.

export LD_LIBRARY_PATH="/usr/local/lib/:$LD_LIBRARY_PATH"

Compile Program with the following command:

clang++ -L/usr/local/lib -l mlpack -l armadillo -l boost_serialization -l boost_program_options -I/usr/local/include KMeans.cpp

Run:

The program takes two command line arguments

• features.csv - input features
• responses.csv - predictors

./a.out <features.csv> <responses.csv>

Note:

• KMeans clustering is an unsupervised learning algorithm, i.e it does not require any labels. However, in this example we use responses to calculate the accuracy of a model.
• The code can be slightly altered to work without the labels and return only the cluster assignments
• There is no train/test split hapenning here. All of the data is used to identify clusters and for each point, cluster assignments are calculated.