Research On Several Improved Density Peak Clustering Algorithms And Their Applications

As an invisible asset in all industries in the new era,data plays a pivotal role in helping industries to transform and upgrade and expand their business.In the context of the current era,digitalization,digital and intelligence,intelligence have undoubtedly become the development opportunity to achieve self-breakthrough in various fields,and data is the core kinetic energy to drive development.Clustering has become an important data analysis method in order to deeply explore the hidden dividends in data resources.Meanwhile,Density Peak Clustering Algorithm(DPC)has been studied and considered by domestic and foreign experts and scholars from different perspectives and perspectives with its own unique advantages.However,the original density peak clustering still has many shortcomings that restrict the improvement of clustering accuracy and clustering performance.Firstly,the cut-off distance relies on human setting.Secondly,the existing local density calculation method is not reasonable enough.And thirdly,there are drawbacks in the data sample distance measurement method which is based on simple geometric distance.Based on this,this paper integrates various theoretical ideas in the original density peak clustering and devotes to the dual improvement of clustering performance and clustering accuracy.The main research contents of this paper are as follows.(1)In order to eliminate the influence of the inadequacy of the existing local density calculation method on the results,this paper is inspired by the research of related scholars and introduces the theory of Laplace distribution in the original density peak clustering,and proposes the Density Peak Clustering Algorithm Based on Laplace Distribution(LPDPC).The algorithm uses the Laplace probability density function to improve the original local density measure,and integrates the global and local distribution of data points in the sample space,which can reasonably solve the problems of unreasonable density ranking and inappropriate distribution of data points in the original density peak clustering algorithm.(2)For how to determine the reasonable cut-off distance value adaptively this paper proposes the density peak clustering algorithm based on bat optimization(BA-DPC).The algorithm introduces bat optimization theory in density peak clustering,and the bat optimization algorithm adaptively scans the parameter space to find the optimal values in the solution space through the variation of pulse frequency,loudness and flight speed,and then realizes the clustering analysis process of the algorithm.The simulation experimental results show that the clustering results of the bat-optimized density-peak clustering algorithm are closer to the real class numbers of the experimental data set,which improves the clustering performance of the original algorithm.(3)In order to eliminate the effect of dimensionality on the density peak clustering algorithm,this paper optimizes the distance measure between data sample points and proposes the Density Peak Clustering Algorithm for Optimization Distance Strategy(ODS-DPC).Inspired by the similarity measure of high-dimensional data,ODS-DPC integrates multiple theories to construct an adaptive distance measure strategy that can be used for data points of different dimensions.Simulation experimental results demonstrate that the improved density-peak clustering algorithm reduces the dimensional impact of high-dimensional data and achieves a better clustering analysis process.