An Improved DBscan Clustering Algorithm

An improved DBscan clustering algorithm is proposed. The improved algorithm based on the following two points:(1) Due to Core Point that is selected by randomly based DBscan algorithm leads to the disadvantage of large computation,and puts forward a method of selecting Core Point based the farthest distance and points in ε distance are more thanMinptspoints. (2) Because parameters of εandMinpts are global uniqueness leads to shortcomings of poor of clustering quality, puts forward a method of Second Clustering, Calculating the distance between cluster center and the noise points have been wrongly selected, the noise points shall be inserted the nearest cluster. At the same time, the quality of clustering is measured by using the silhouette coefficient. The experimental results show that:Compared with the original DBscan clustering algorithm the algorithm has better performance efficiency and clustering quality.