Research on a New Density Clustering Algorithm Based on MapReduce

Abstract

The empirical solution parameters for the Density-Based Spatial Clustering of Applications with Noise(DBSCAN) resulted in poor Clustering effect and low execution efficiency, An adaptive DBSCAN algorithm based on genetic algorithm and MapReduce programming framework is proposed. The genetic algorithm (minPts) and scanning radius size (Eps) optimized intensive interval threshold, at the same time, combined with the similarities and differences of data sets using the Hadoop cluster parallel computing ability of two specifications, the data is reasonable of serialization, finally realizes the adaptive parallel clustering efficiently. Experimental results show that the improved algorithm (GA) - DBSCANMR when dealing with the data set of magnitude 3 times execution efficiency is improved DBSCAN algorithm, clustering quality improved by 10%, and this trend increases as the amount of data, provides a more precise threshold DBSCAN algorithm to determine the implementation of the method.