A Massive Data Reduction Scheme Based on Improved Hausdorff Distance

Abstract

With the rapid development of information technology, present quite a few massive data, need a lot of time to analyze these complex data, but the current reduction methods also have its drawbacks. In this paper, combine with Hausdorff distance superiority in measuring data sets similarity, propose an improved data reduction method based on local Hausdorff distance, utilize an improved K-NN retrieval algorithm to equably divide the original data sets, obtain a number of sub-data sets, carry out the FCM clustering operation to determine the center point of each category, perform reduction on each sub-data sets, merge each sub-reduction data sets to get the final reduction data sets. The experimental results shown that this method doesn’t rely on experience and knowledge, only need to consider the cases’ distribution in data sets while performing reduction, has a certain amount of universality for case-based data reduction.