Data Mining and Knowledge Engineering

A. Gnanabaskaran ¹, K. Duraiswamy ¹

Affiliations
1 K.S. Rangasamy College of Technology, Tiruchengode-637215, IN

Abstract

Clustering low dimensional spatial data for traffic risk analysis has been major issue due to sparsity of data points. Most of mthe clustering algorithm becomes inefficient if the required distance similarity measure is computed for low dimensional spatial space of high dimensional data with sparsity of data point along different dimensions. objective of this study were to contribute the complexity of projecting clusters for traffic risk analysis, (i) There is no support for minimizing the number of dimensions on spatial space in order to reduce the searching time (ii) Comparison of computation time of HARP, Proclus, Doc, FastDoc, SSPC algorithms. During the first phase the satellite captured still images for different dimensions network are enhanced and this images are given as input to second phase spatial attribute relevance analysis for detecting dense and sparse regions after detecting dense and sparse regions the algorithm employees pruning technique to reduce the search space by taking only dense traffic regions and eliminating sparse traffic region and during third phase K-means algorithm is employed to project the clusters on different spatial dimensions. As per the Results first we showed that various projecting clustering algorithm on spatial space becomes inefficient if the number of dimensions increases. The new scheme proposed reduces the spatial dimension space so that it reduces the computation time and finally the result is compared with HARP, Proclus, Doc, FastDoc, SSPC. The algorithms produces acceptable results when the average cluster dimensionality is greater then 10%. Hence as for as Conclusion the findings suggested the overhead reasonably minimized and using simulations, we investigated the efficiency of our schemes in supporting low dimensional spatial clustering for traffic risk analysis.

Keywords

Data Mining, Clustering, Low Dimensions, Projected Clustering.

Full Text

     

B. Jayanthi ¹, K. Duraiswamy ²

Affiliations
1 P.G.Department of Computer Science, Kongu Arts and Science College, Erode, Tamilnadu, IN
2 K.S. Rangasamy College of Technology, Tiruchengode, Tamilnadu, IN

Abstract

In this paper an algorithm is proposed for mining frequent itemsets. This paper proposes a new framework to generate frequent Itemsets/Patterns. First, a partitioning technique is used to divide a transaction database TDB into n non-overlapping partitions. Second we use fp-graph model to discover frequent itemsets for each partition. Example illustrating the proposed approach is given. The characteristics of the algorithm are discussed.

Keywords

Data Mining, Frequent Patterns, Frequent Itemset, Partitioning Technique, FP-Graph, Association Rule.

Full Text