Density-based semi-supervised clustering

Abstract

Semi-supervised clustering methods guide the data partitioning and grouping process by exploiting background knowledge, among else in the form of constraints. In this study, we propose a semi-supervised density-based clustering method. Density-based algorithms are traditionally used in applications, where the anticipated groups are expected to assume non-spherical shapes and/or differ in cardinality or density. Many such applications, among else those on GIS, lend themselves to constraint-based clustering, because there is a priori knowledge on the group membership of some records. In fact, constraints might be the only way to prevent the formation of clusters that do not conform to the applications' semantics. For example, geographical objects, e.g. houses, separated by a borderline or a river may not be assigned to the same cluster, independently of their physical proximity. We first provide an overview of constraint-based clustering for different families of clustering algorithms. Then, we concentrate on the density-based algorithms' family and select the algorithm DBSCAN, which we enhance with Must-Link and Cannot-Link constraints. Our enhancement is seamless: we allow DBSCAN to build temporary clusters, which Responsible editor: Eamonn Keogh. 123 346 C. Ruiz et al.we then split or merge according to the constraints. Our experiments on synthetic and real datasets show that our approach improves the performance of the algorithm.

Semi-Supervised Kernel Mean Shift Clustering

“…The initial kernel matrix is computed using a Gaussian kernel (33) for all the methods. We hand-picked the scale parameter σ for SSKK and E2CP from a wide range of values such that their final clustering performance on each data set was maximized.…”

“…SSDBSCAN [26] is a semi-supervised variant of DBSCAN that explicitly uses the labels of a few points to determine the neighborhood parameters. C-DBSCAN [33] performs a hierarchical density based clustering while enforcing the must-link and cannot-link constraints.…”

“…From (33) and (11) percentile of distances between all pairs of points in the kernel space. We select the scale parameter σ such that, in the initial kernel space, distances between must-link points are small while those between cannot-link points are large.…”

Density-Based Clustering with Geographical Background Constraints Using a Semantic Expression Model

“…However, the method has many shortcomings in supporting the use of geographical background knowledge for the formulation of constraints. Ruiz et al developed the C-DBSCAN (density-Based clustering with constraints) [8] method, which offers improved performance over DBSCAN with a small number of constraints. However, C-DBSCAN cannot handle real geographical constraints, and is lack of semantics.…”

Markov Prediction based on Semi-supervised Kernel Fuzzy Clustering

“…In recent years, machine learning and semi-supervised clustering algorithm in data mining got the attention of many scholars, especially in industrial applications, for the vague and incomplete data that can't be directly referenced in the industrial scene, often spend a lot of manpower and time cost [1,2] . Semi-supervised clustering used the prior knowledge implied the data to integrate these unlabeled data in order to realize the clustering.…”