A Novel Density Based Clustering Algorithm by Incorporating Mahalanobis Distance

Abstract

Data clustering is one of the active research areas, which aims to group related data together. The process of data clustering improves the data organization and enhances the user experience as well. For this sake, several clustering algorithms are proposed in the literature. However, a constant demand for a better clustering algorithm is still a basic requirement. Understanding the necessity, this paper proposes a density based clustering algorithm which is based on Density Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. The main drawback of DBSCAN algorithm is it requires two important parameters as initial input. It is really difficult to fix the values for these parameters, as it requires some prior knowledge about the dataset. This requirement is overthrown by the proposed clustering algorithm by selecting the parameters automatically. The automated selection of parameters is achieved by analysing the dataset and it varies from dataset to dataset. This way of parameter selection improves the quality of service and produce effective clusters. The experimental results show that the proposed approach outperforms the DBSCAN algorithm in terms of purity, F-measure and entropy.

Modeling Vessel Behaviours by Clustering AIS Data Using Optimized DBSCAN

“…Various methods have also been suggested to increase the algorithm's computational efficiency when applied to large databases [46][47][48][49][50]. Additionally, various methods presenting new clustering conceptions to DBSCAN can be found in [19,23,24,51].…”

“…This leads to unreliable results when applying the traditional DBSCAN method to real AIS data without optimization. Furthermore, the traditional DBSCAN method, based on the Euclidean distance metric, can face challenges with data that have complex shapes and distributions [23,24]. Thus, novel distance metrics need to be proposed to optimize DBSCAN performance.…”

“…Most of the existing optimizations are designed for clustering two-dimensional spatial data (i.e., x, y). When the data dimensions grow and the Mahalanobis distance metric is used, the distribution of the dataset becomes different [13,23,24]. Therefore, the existing adaptive parameter method needs to be modified prior to its application to the enhanced DBSCAN method.…”

Automated Digital Image Clustering Algorithm Based on Colour Distance and IDBSCAN

“…The unsupervised learning is quite popular in different applications of data mining, digital image processing, pattern recognition and so on [1][2][3][4]. Basically, the clustering algorithms can be classified into four categories such as partitional, hierarchical, fuzzy, density and grid based clustering [5,6]. Out of all these clustering algorithms, this work focuses on density based clustering, as this kind of clustering performs well in forming arbitrary shaped clusters and the spatial relationship between the data entities is also maintained.…”

“…DBSCAN is one of the most promising density based clustering algorithm, which can withstand noise, shapes and densities. However, the functionality of the DBSCAN algorithm completely relies on two parameters, which are (epsilon) and [5]. Epsilon is the radius from a corresponding pixel, which includes neighbouring pixels.…”