Abstract:
Traditional DBSCAN algorithm usually set the threshold(min
Pts,
Eps) depending on empirical value,and merely fit for single-density data sets,in allusion to the shortage,the article puts forward a new adaptive DBSCAN algorithm which based on Gaussian distribution and can be used to multi-density data sets.The new algorithm can generate proper threshold according to the data set characteristics.Firstly,to get the min
Pts accornding to the max CEI value.Then,to confirm the number of
Eps according to the curve density level,so that to get
Eps data by the Gaussian distribution law.At last,to make clustering for the data set with the min
Pts and
Eps value.Also,to apply the new algorithm and the traditional DBSCAN algorithm to single-density and multi-density data sets,the results show that the new algorithm is more efficient.