A Combined System for Regionalization in Spatial Data Mining Based on Fuzzy C-Means Algorithm with Gravitational Search
The proposed new hybrid approach for data clustering is achieved by initially exploiting spatial fuzzy c-means for clustering the vertex into homogeneous regions. Further to improve the fuzzy c-means with its achievement in segmentation, we make use of gr
- PDF / 250,555 Bytes
- 8 Pages / 439.37 x 666.142 pts Page_size
- 8 Downloads / 166 Views
Abstract The proposed new hybrid approach for data clustering is achieved by initially exploiting spatial fuzzy c-means for clustering the vertex into homogeneous regions. Further to improve the fuzzy c-means with its achievement in segmentation, we make use of gravitational search algorithm which is inspired by Newton’s rule of gravity. In this paper, a modified modularity measure to optimize the cluster is presented. The technique is evaluated under standard metrics of accuracy, sensitivity, specificity, Map, RMSE and MAD. From the results, we can infer that the proposed technique has obtained good results.
⋅
⋅
Keywords Spatial data mining Clustering search algorithm Regionalization Metrics
⋅
⋅
Fuzzy c-means
⋅
Gravitational
1 Introduction In spatial data mining, knowledge discovery denotes the unearthing of implied, previously unknown and interesting knowledge from spatial databases. The important task in spatial data mining is spatial clustering. It aims to group similar spatial objects into classes or clusters. The objects in a cluster have high similarity in estimation to one another and are distinct to objects in other clusters [1]. A prime application area for spatial clustering algorithms in geography is towards social and economical issue. In the scope a classical methodical problem of social geography is “regionalisation”. In general it is known as a classification procedure applied to spatial objects with an area representation. It groups them into contiguous regions of homogeneous nature [2, 3]. The main objective of regionalization is to find Ananthi Sheshasaayee (✉) Quaid-E-Millath Goverment College for Women (Autonomous), Chennai, Tamil Nadu, India e-mail: [email protected] D. Sridevi Computer Science Department, Sri Chandrasekharendra Saraswathi Viswa Maha Vidyalaya, Kanchipuram, Tamil Nadu, India © Springer Nature Singapore Pte Ltd. 2017 S.C. Satapathy et al. (eds.), Proceedings of the 5th International Conference on Frontiers in Intelligent Computing: Theory and Applications, Advances in Intelligent Systems and Computing 516, DOI 10.1007/978-981-10-3156-4_54
517
518
Ananthi Sheshasaayee and D. Sridevi
spatial regions of not a specific shape (arbitrary) with a homogeneous internal distribution of non-spatial variables with respect to compactness and density [4]. In this paper, we suggest a new hybrid approach for data clustering making use of FCM and gravitational search optimization with the results obtained for the different clusters.
2 Literature Survey In the literature survey, several methods have been proposed for the Regionalization in spatial data mining. Among the most recently published works are those presented as follows: Niesterowicz and Stepinski [5] has explained the regionalization of multi-categorical landscape. The land cover pattern is based on the principle of machine vision rather than clustering of landscape metrics. The NLCD 2006 shows spatially varying pattern of Land Use/Land Categories of maps. Using an LULC map as an input of their method locates and ma
Data Loading...
