Data gravitation based classification
Information Sciences: an International Journal
A DGC-based data classification method used for abnormal network intrusion detection
ICONIP'06 Proceedings of the 13th international conference on Neural information processing - Volume Part III
Gravitation based classification
Information Sciences: an International Journal
| Hi-index | 0.00 |
In this paper, we present continuous research on dataanalysis based on our previous work on the shrinking approach.Shrinking[A shrinking-based approach for multi-dimensional data analysis] is a novel data preprocessing technique which optimizes the inner structure of data inspiredby the Newton's Universal Law of Gravitation[The laws of physics] in the realworld. It can be applied in many data mining fields. Followingour previous work on the shrinking method for multi-dimensionaldata analysis in full data space, we propose ashrinking-based dimension reduction approach which tendsto solve the dimension reduction problem from a new perspective.In this approach data are moved along the directionof the density gradient, thus making the inner structureof data more prominent. It is conducted on a sequence ofgrids with different cell sizes. Dimension reduction processis performed based on the difference of the data distributionprojected on each dimension before and after the data-shrinkingprocess. Those dimensions with dramatic variationof data distribution through the data-shrinking processare selected as good dimension candidates for furtherdata analysis. This approach can assist to improve the performanceof existing data analysis approaches. We demonstratehow this shrinking-based dimension reduction approachaffects the clustering results of well known algorithms.