Efficient Constrained K-center Clustering with Background Knowledge

Data preprocessing

• Following the label for the dataset, we add CL and ML constraints on the pure dataset by label (which constructs the disjoint CL & ML constraints) to be input dataset for the main algorithm.
• The input of the program can be pure constraint sets for CL & ML.

Run the code

• Java version: 17.0.1
• Cplex: IBM CPLEX2211 https://www.ibm.com/products/ilog-cplex-optimization-studio
• Run OuPut.java for the experimental evaluation
• To simplify the code, we hard-code the following parameters:

Input [ML-CL-k-Center/code/addConstraints/OutPut.java]

- set the file path at the _inputFilename_ (and the pure dataset needs to separate attribute values with commas);
- input the parameter k: number of clusters, d: dimension, markPosition: label position

Output [ML-CL-k-Center/code/addConstraints/OutPut.java]

- set the file path at the _outputFilename_;

Plot the output

• Use the Cost, NMI, RI and runtime to calculate the agreement degree between an algorithm's clustering result and its labels.