Improvement Of Clustering Algorithm And Its Application In Sensor System's Identification

Data mining is a new direction of data analysis.The technology can find the internal relationship what are not easy to detected between the data sets,and brings new solutions for commercial,industrial,medical and other decisions.The research on clustering mining has become increasingly important.How to reduce the time complexity and improve the optimization accuracy,and how to find the right scene to adapt which become the main objective of the current study about reducing the cost of consumption.This dissertation introduces the background knowledge of data mining and the clustering algorithm.Then improve the algorithm based on density peak and grid.In the end,use the improved algorithm to identificate the sensor nonlinear system.The main work and achievements of this dissertation are as follows:1.According to the clustering algorithm that clustering by fast search and find of density peaks(DPC),it's density calculation using Euclidean distance to measure the similarity between two points.In this dissertation,we propose a new idea of combining DPC algorithm with grid algorithm,and use the grid division technology to convert all point objects into grid objects.Especially the improved algorithm shows good performance when the amount of data is very large.2.According to the DPC algorithm's defect which determining the number of cluster centers needs to manually select through decision diagram,this dissertation adopts an improved criterion,which can automatically determine the clustering center points and number of clusters.The improvement of this step provides the base of industrial production automation,reducing the process of human decision-making.3.According to the grid edge's points and noise,to remove them by using the principle of grid's similarity.The adjacent grids but do not belong to the same cluster,which comparing a point's similarity between it's grid and it's adjacent grid.4.According to the identification of nonlinear system sensor fault and the influence of sensor data on the work in the state system,to determine the results of parameter identificationof nonlinear system model,analyzing the input and output data with the improved clustering algorithm.The MATLAB simulation results show that this algorithm effects well in sensor's nonlinear system identification and fault identification.