Research Of Data Field In Cluster Analysis

During the process of the development of many fields in information era, that will produce large amounts of data information, which may contain some hidden pattern depending on data mining to find the rule. Clustering analysis is an important method about data mining technology research, which has been get more and more attention in the field of study.The core of clustering analysis is the similarity of metric data objects. In geometric sense, the similarity is defined by the distance between the data objects. The distance is larger, the similarity is lower; the distance is smaller, the similarity is higher; Data field is an abstract of "field" concept in physics, which Illustrates an energy relation between the data. The high energy intensity concentrated area in data field is the place of the populated data objects, the similarity of the places of populated data objects is higher, which will polymerize a kind of class cluster. The data field can be used in the clustering analysis, which will be a possibility to use a data field for clustering analysis.Since the study of data field clustering is just emerging, it is not very prefect in many respects. For example, the uncertainty of selecting potential functions, no fixed measure of parameter selection, no very good support to high-dimensional data clustering and the method is too single and the application is inadequate. In view of great practical value of the cluster analysis and solid theoretical ideas of data field, and according to the research of the data field clustering technologies and development trends, this paper shows the importance and necessity of the data field clustering analysis.Through the establishment and study of the different models in data field, to choose the better data field model for clustering analysis, to make a comprehensive research about parameters of the radiation intensity. Studies have shown that the selection model of the radiation field and Radiation brightness parameter adjusting in the golden point will achieve the best clustering results.To test the robustness of data field clustering algorithm, a clustering analysis is carried by constructing complex data sets. The experimental results show that the clustering analysis algorithm can effectively identify the class cluster of any shape, and it has better recognition effect to the hidden class cluster in the noise, which shows its advantages of resistance to the noise.To the high dimensional data sets, we have made a test on several typical of the UCI data sets by reducing the dimension, and the results are compared with other traditional clustering algorithm results. The experimental results show that the clustering algorithm in the higher dimension of data sets have been achieved a better clustering result.Based on the characteristics of the data field itself, feature points of face image are extracted using the data field in this paper, and a new method of face recognition is proposed for a inquisitive attempt, and this method is a good reference for face recognition.