Out Lier Detection and Clustering Analysis in Data Stream Classification

Neethu S, Sajni Nirmal

Abstract: In social network stream, we can detect anomalies and emerging topics based on links between the users that are generated dynamically. In the rapid growth of social network, discovering emerging topics and classification of data is most important challenging issues. For the emerging topic detection purpose to propose a new method in the area of streaming data. Here, Dynamic Threshold Optimization algorithm is used to detect anomalies in streaming data. Anomaly detection refers to detecting clusters or objects in a given data set that conform to an established abnormal behavior. This classes and object are called anomalies or outliers that are critical information in several application domains. To find anomalies in social streams by using various clustering methods. This paper also implements the data mining technique like text clustering methods to clustering the dataset which contains medical records of patients. Here, we applied Hierarchical text clustering methods like C- Mean, Hierarchical Agglomerative clustering, and Single linkage algorithms are used for clustering and classification of the dataset. LKC algorithm and compatibility is introduced to provide privacy preservation. The performance analysis carried out by different clustering algorithms posses processing of data in stream and dataset.

Keywords: Dynamic threshold optimization, Outlier detection, Privacy preservation, Text clustering methods

How to Cite?: Neethu S, Sajni Nirmal, "Out Lier Detection and Clustering Analysis in Data Stream Classification", Volume 5 Issue 6, June 2016, International Journal of Science and Research (IJSR), Pages: 1207-1210, https://www.ijsr.net/getabstract.php?paperid=NOV164400, DOI: https://dx.doi.org/10.21275/NOV164400