UNCERTAIN DATA CLUSTERING IN PEER TO PEER NETWORKS

Abstract

A sensational rise of various applications and businesses has led to the rise of data being collected, it is getting harder to store this data on a single machine. It has become more feasible to manage dispersed data sets and organizations. Recently a new technique has been introduced called distributed data mining (DDM). It is becoming popular due to the fact that analysing huge data sets present at different locations, applying traditional centralized techniques could be inefficient. That is the reason distributed data mining methods have become popular. Perhaps the main hurdle of data mining is to perform data clustering. It is quite possible that data produced in distributed environments contain noise or obsolete information, this type of data is termed as uncertain data. This type of data is difficult to cluster and consequently it becomes difficult to make any business decision or inference. Uncertain data clustering has been perceived as a fundamental step in the analysis of data mining, or in other words it has become a part of pre-processing. Many centralized clustering calculations have been modified by characterizing new estimations. While many clustering calculations have been introduced for certain and uncertain data sets, there is need of productive calculations for distributed data sets. A modified hierarchical clustering algorithm is proposed for uncertain dataset to improve timing in execution of the process. This algorithm does not require any pre-specified number of clusters. Efficiency of the algorithm is worth noting and is tested with real world data sets.