Peak density algorithm based on KD-tree optimization

In view of the high time complexity density peak algorithm, it is necessary to manually confirm clustering center according decision graph. A peaking algorithm KT-DPC based on kd-tree optimization proposed. The defines local p through K-nearest neighbor and uses accelerate distance 8. addition, in confirmation stage center, a strategy (C2BD, difference) difference between adjacent y calculated by arranging y=p*S ascending order, boundary non-clustering found change difference. This method automatically confirms avoiding problems strong subjectivity insufficient accuracy caused confirming center. Experiments multiple UCI public data sets show that running under low-dimensional are better than traditional DPC algorithms, KNN-DPC other improved algorithms.