This repository presents the C++ implementation of the ILP models described in the article "Qualitative data clustering: a new Integer Linear Programming model".

Two header libraries are avaiable:

• ILP.h: creates and runs the ILP models on the IBM's CPLEX Optimizer™

• LPCreator.h: creates the ILP models on CPLEX LP format, so it can be run on any compatible solver.

• Example usage of ILP
• Example usage of LPCreator

ILP.h provides the class

ILP(std::string dataset)

where

• dataset - name of the input dataset file

ILP class has a method ILP::execute(int modelType) that receives as parameter the type of ILP model you wish to run:

• 0 - Runs the complete model (GM)
• 1 - Runs model GM1
• 2 - Runs model GM2
• 3 - Runs model GM3

The following code excerpt shows how to run the complete model (GM) on the dataset 1-Lenses.txt

#include "ILP.h"

using namespace std;
using namespace qualitativeclustering;

int main()
{
    ILP ilp("data/1-Lenses.txt");
    cout << ilp.execute(0);
}

The output will be

Objective = 72
Constraints = 6072
Time Solve = 0.446455
Groups = 0 1 0 1 2 2 3 3 0 1 0 1 2 2 3 3 0 1 0 1 2 2 3 3

You must have IBM's CPLEX Optimizer™ 12> installed on your system to compile this code excerpt.

Let's suppose you are on Mac and <CPLEX_DIR> is the directory where CPLEX is installed on your system, the following command will compile this example:

g++ main.cpp -o main -I/<CPLEX_DIR>/cplex/include 
                     -I/<CPLEX_DIR>/concert/include
                     -DIL_STD 
                     -L/<CPLEX_DIR>/cplex/lib/x86-64_osx/static_pic/ 
                     -L/<CPLEX_DIR>/concert/lib/x86-64_osx/static_pic/ 
                     -lilocplex -lconcert -lcplex -lm -lpthread -std=c++11

For more information regarding setting up CPLEX on Linux/Mac/Windows:

• Compiling on LINUX/MAC: You can follow the rules defined in Setting up CPLEX on GNU/Linux to compile it on Linux/Mac.

• Compiling on WINDOWS: The rules defined in Setting up CPLEX on Windows can be used to compile it on Windows environment.

LPCreator.h provides the method to create ILP models in LP format

void createLP(std::string dataset, std::string LPfilename, int modelType)

where

• dataset - name of the input dataset file
• LPfilename - name of the output LP file
• modeltype - type of model to create:
  • 0 - Create the complete model (GM)
  • 1 - Create GM1 model
  • 2 - Create GM2 model
  • 3 - Create GM3 model

The following code excerpt shows how to create an LP file that represents the complete model (GM) for the dataset 1-Lenses.txt:

#include "LPCreator.h"

using namespace std;
using namespace qualitativeclustering;

int main()
{
    createLP("data/1-Lenses.txt", "1-Lenses-GM.lp", 0);
}

The output will be a file in CPLEX LP format that can be used on any compatible solver.

LPCreator.h is a single header-only library. Hence, to compile this code excerpt just type on Linux/Mac:

g++ main.cpp -o main -std=c++11