This is the semester long project for a graduate class in cluster analysis at my university. The code remains mostly unchanged since its final submission. However, some changes have been made to protect my privacy and improve the code quality.
ℹ️ Note: This project uses zero dependencies by design.
It was a project requirement that no libraries could be used, except for the language's standard library. There is also no unit testing, as I did not have enough time to write them. I relied on hints given by project documents and the grade given to me by the professor to know if the code was working correctly.
Building this project is simple. Just follow these steps:
- Install Git if not done so already.
- Install the JDK. Builds of the OpenJDK can be found here.
- Open a terminal of your choice and run the commands below.
git clone https://github.com/whirvis/clusteral
cd clusteral
chmod +x ./gradlew # unix only
./gradlew build
# install to local Maven repository if desired
./gradlew publishToMavenLocalBefore running the clustering program, you will need to have a compatible dataset file. The program expects each file to have a header line, and then subsequent lines for each data point. An example is listed below.
250 7 3
0.63 0.47 0.48 0.51 0.82 0.84 0
0.23 0.48 0.48 0.59 0.88 0.89 0
0.34 0.49 0.48 0.58 0.85 0.80 0
0.43 0.40 0.48 0.58 0.75 0.78 0
0.46 0.61 0.48 0.48 0.86 0.87 0
0.27 0.35 0.48 0.51 0.77 0.79 0
0.52 0.39 0.48 0.65 0.71 0.73 0
0.29 0.47 0.48 0.71 0.65 0.69 0
0.55 0.47 0.48 0.57 0.78 0.80 0
0.12 0.67 0.48 0.74 0.58 0.63 0
(240 more points...)Note the contents of the first line. The first number is how many data points there are, the second number is how many dimensions each point has, and the third number is the number of true clusters. Keep in mind that the true cluster is also one of the axes! It is always the last axis of each point.
There is a main method in ClusteringProgram. It should be executable out
of the box with valid arguments. Parameters are listed in ClusteringParams.
An example of valid CLI arguments are given below.
java -jar clusteral.jar \
dataset.txt # F: name of the data file \
3 # K: number of clusters \
1000 # I: maximum number of iterations \
0.001 # T: convergence threshold \
100 # R: number of runs \
random-selection # M: K-means initialization method \
min-max # N: normalization type \
calinski-harabasz # V: name of the cluster validator \
average # D: diameter calculation method \
single-linkage # L: linkage method, if necessary (optional) \
true # C: random centroids on multiple nearest (optional) \
stdout # W: where to write program output (optional)Both classes are in the io.whirvis.edu.clustering.cli package of the cli
module.