Virtual Subconcept Drift Detection in Discrete Streaming Data Using Probabilistic Graphical Models

R. Cabañas, A. Cano, M. Gómez-Olmedo, A. Masegosa, S. Moral

Paper presented at IPMU 2018

This repository contains the required material for the experimentation given in the paper (see full-text). In particular, it is organised as follows:

• R: this folder contains all the R scripts for generating the data sets, modifying them, and creating the graphs with the results.

• datasets: this folder contains the data sets used in the experimentation (i.e., the synthetic and intrusion data sets) which are provided in multiple .arff files.

• java: maven project for running the experiments. This uses AMIDST 0.6.2. The code implementing the concept drift detector is given in this folder. For easily running the experiments shown in the paper, see the section below.

Citation:

@InProceedings{CabanasIPMU18virtualsubconcept,
author="Caba{\~{n}}as, Rafael and Cano, Andr{\'e}s and G{\'o}mez-Olmedo, Manuel and Masegosa, Andr{\'e}s R. and Moral, Seraf{\'i}n",
title="Virtual Subconcept Drift Detection in Discrete Data Using Probabilistic Graphical Models",
booktitle="Information Processing and Management of Uncertainty in Knowledge-Based Systems. Applications",
year="2018",
publisher="Springer International Publishing",
address="Cham",
pages="616--628",
isbn="978-3-319-91479-4"
}

Running the paper experiments

First, download the example project code:

$ git clone https://github.com/PGMLabSpain/2017-CDdiscrete-Code.git

Enter in the downloaded java folder:

$ cd 2017-CDdiscrete-Code/java/

The code for running the experiments is in the file ./src/main/java/impu_run.java. You can modify it as you want.

Compile and build the package:

$ mvn clean package

Finally, run the code example previously mentioned:

$ java -cp target/example-project-full.jar ipmu_run

Requirements: git, maven, Java8