Chiheb-Eddine Ben N'cir

Identification of non-disjoint clusters with small and parameterizable overlaps

Identification of non-disjoint groups in unlabeled data sets is an important issue in clustering. Many real life applications require to find overlapping clusters in order to fit the data set structures such as clustering of films where each film can have different genres. This paper presents an overlapping k-means method refereed as Restricted-OKM (Restricted Overlapping k-means) that generalizes the well known k-means algorithm to detect overlapping clusters. The proposed method produces restricted overlapping boundaries between clusters and improves clustering accuracy to make the method adapted for clustering data with small overlaps. The proposed method is extended to control sizes of overlaps between clusters with respect to user expectations. Experiments, performed on overlapping data sets, show that proposed methods outperform OKM (Overlapping k-means) and fuzzy c-means in terms of clustering accuracy and produce clusters with small overlapping boundaries.

Overlapping Patterns Recognition with Linear and Non-Linear Separations using Positive Definite Kernels

The detection of overlapping patterns in unlabeled data sets referred as overlapping clustering is an important issue in data mining. In real life applications, overlapping clustering algorithm should be able to detect clusters with linear and non-linear separations between clusters. We propose in this paper an overlapping clustering method based k-means algorithm using positive definite kernel. The proposed method is well adapted for clustering multi label data with linear and non linear separations between clusters. Experiments, performed on overlapping data sets, show the ability of the proposed method to detect clusters with complex and non linear boundaries. Empirical results obtained with the proposed method outperforms existing overlapping methods.

Kernel-Based Methods to Identify Overlapping Clusters with Linear and Nonlinear Boundaries

Detecting overlapping structures and identifying non-linearly-separable clusters with complex shapes are two major issues in clustering. This paper presents two kernel based methods that produce overlapping clusters with both linear and nonlinear boundaries. To improve separability of input patterns, we used for both methods Mercer kernel technique. First, we propose Kernel Overlapping K-means I (KOKMI), a centroid based method, generalizing kernel K-means to produce nondisjoint clusters with nonlinear separations. Second, we propose Kernel Overlapping K-means II (KOKMII), a medoid based method improving the previous method in terms of efficiency and complexity. Experiments performed on non-linearly-separable and real multi-labeled data sets show that proposed learning methods outperform the existing ones.

Parallel clustering method for non-disjoint partitioning of large-scale data based on spark framework

Clustering large scale data has become an important challenge which motivates several recent works. While the emphasis has been on the organization of massive data into disjoint groups, this work considers the identification of non-disjoint groups rather than the disjoint ones. In this setting, it is possible for data object to belong simultaneously to several groups since many real-world applications of clustering require non-disjoint partitioning to fit data structures. For this purpose, we propose the Parallel Overlapping k-means method (POKM) which is able to perform parallel clustering processes leading to non-disjoint partitioning of data. The proposed method is implemented within Spark framework to ensure the distribution of works over the different computation nodes. Experiments which we have performed on simulated and real-world multi-labeled datasets shows both faster execution times and high quality of clustering compared to existing methods.

Non-disjoint Cluster Analysis with Non-uniform Density

Non-disjoint clustering, also referred to as overlapping clustering, is a challenging issue in clustering which allows an observation to belong to more than one cluster. Several overlapping methods were proposed to solve this issue. Although the effectiveness of these methods to build non-disjoint partitioning, they usually fail when clusters have different densities. In order to detect overlapping clusters with uneven densities, we propose two clustering methods based on a new optimized criterion that incorporates the distance variation in a cluster to regularize the distance between a data point and the cluster representative. Experiments performed on simulated data and real world benchmarks show that proposed methods have better performance, compared to existing ones, when clusters have different densities.

Overlap regulation for additive overlapping clustering methods

Overlapping Clustering is an important technique in machine learning which aims to organize data into a set of non-disjoint groups rather than the disjoint one which is the case of conventional clustering methods. Several machine learning applications require that data object be assigned to one or several groups resulting in non-disjoint partitioning of data such as document clustering where each document can discuss one or many topics and then must be assigned to one or several groups. This paper presents a new partitional overlapping clustering method based on the additive model of overlaps. Compared to existing methods which build clusters with fixed size of overlaps, the proposed method gives users the ability to regulate this size. Experiments performed on simulated and real datasets show the performance of the proposed regulation principle to control the size of overlaps among groups.

Overlapping Clustering with Outliers Detection

Detecting overlapping groups is an important challenge in clustering offering relevant solutions for many applications domains. Recently, Parameterized R-OKM method was defined as an extension of OKM to control overlapping boundaries between clusters. However, the performance of both, OKM and Parameterized R-OKM is considerably reduced when data contain outliers. The presence of outliers affects the resulting clusters and yields to clusters which do not fit the true structure of data. In order to improve the existing methods, we propose a robust method able to detect relevant overlapping clusters with outliers identification.