A Novel Approach for Data Mining Clustering Technique Using NeuralGas Algorithm

Efficient Privacy preserving association rule mining has emerged as a latest research issue. In this thesis work, existing algorithms, Increase Support of Left and Decrease Support of Right are implemented successfully on the real data for Privacy Preserving Association Rule Mining. In order to hide an association rule, a hybrid algorithm is proposed which is based on two previous existing algorithms ISL and DSR. K-Mean, Neural gas Cluster Algorithm with Number of cluster in this algorithm, first we decrease support of right hand side of the rule in a rule where item to be hide is in right side for Experimental work, we have used a real time database of Doctor Patient Evaluation from Medical College.

I. INTRODUCTION

In this thesis assuming that only sensitive items are given we implement two existing algorithms, ISL (Increase Support of LHS) [1] and DSR (Decrease Support of RHS) [2], we also compare all four algorithms on the basis of number of database scans and no. of clusters. We proposed Neural Gas algorithm, which can efficiently works on clustering of nonlinearly structured datasets. Compared with several clustering algorithms k-mean algorithm can be less sensitive to initializations due to employing the sequential learning and the neighborhood cooperation scheme. Distortion Sensitive Neural Gas algorithm is also devised to tackle imbalanced clustering issues. Overall results outcome demonstrates the superior performance of Neuralgas Cluster and K-Mean and ISL, DSR Algorithm with Number of Clusters over time in Milliseconds. We also discovered that clustering performances of the methods were dependent on the choice of the parameter. Now we are researching on a new way to adaptively determine suitable parameter values for given clustering tasks.