Research On Adaptive Elastic Network Algorithm For Clustering With New Characteristics

With the advent of the era of big data information,the technical requirements for massive data processing are getting higher and higher.In order to find useful and implicit information in massive data more efficiently,the requirements of cluster analysis technology as a data processing tool are getting higher and higher.Correspondingly,cluster analysis algorithm is the core content of cluster analysis technology.It is also facing increasing demands.At present,many scholars have proposed a number of clustering algorithms.According to the differences or improvements of clustering ideas and rules,conventional algorithms can be divided into five categories,include partitioning clustering,hierarchical clustering,clustering algorithms based on density,grid and model.In addition,some new clustering algorithms have been proposed,including particle-based kernel clustering algorithms and spectral clustering algorithms and many more.However,these clustering algorithms are designed and implemented according to different types and different needs of data sets.They are generally targeted,but they also have limitations and singularity of wide application,that is,they cannot be effectively applied to existing plurals.Clustering scenarios for large data sets.In order to meet the needs of cluster analysis of multivariate and high-dimensional large data sets,this paper proposes a new adaptive clustering elastic network algorithm NAENC.The clustering analysis problem of solving different types of data sets around elastic network is studied.The main research contents include:(1)designing the mathematical model relationship between new data points and elastic nodes;(2)designing adaptive learning ability Dynamic parameter control strategy;(3)adjust the elastic network model structure;(4)design algorithm optimization strategy.At the same time,in order to verify the performance and advantages of the NAENC algorithm proposed in this paper,the artificial random data set of unknown clustering results is tested and compared with the classical partitioning clustering algorithm.In order to verify the truth and reliability of the NAENC algorithm,this paper tests data sets of different sizes and dimensions from the UCI(http://archive.ics.uci.edu/ml/index.php)database for known clustering results,and with the classics.The partitioning clustering algorithms and DBSCAN were compared and analyzed.Through the above experimental tests,the comparison of the test results shows that the SED value of the NAENC algorithm is reduced by about 20%(in this paper,the SED value is lower,the clustering quality is higher).Through comparative analysis,the algorithm clustering results are stable,the network speeds up the convergence speed,greatly improves the clustering quality and clustering efficiency,and saves time and space overhead.In general,the NAENC algorithm proposed in this paper can overcome the shortcomings of traditional clustering algorithms in solving large-scale and high-dimensional data sets,and can effectively avoid such problems as low solution quality,large time overhead,large space overhead,and quality of solution.Algorithmic flaws such as instability and slow network convergence.At the same time,the algorithm can be well applied to the clustering problem of multivariate big data sets.