Research And Application Of Fuzzy Clustering In Hydrology Distribution

Clustering analysis is a branch of Data Ming, and fuzzy cluster is a kind of cluster, many achievements have been proposed, in which, fuzzy C-means clustering has good performance and value. However, the trouble sensitive to the initial cluster center affect the quality and performance of its application. In this dissertation, the research on the trouble is done, the contribution are as following.(1) The problems and the state of data mining and fuzzy clustering are overviewd,and the related concepts and research are introduced.(2) Neural network and simulated annealing algorithm are discussed. Separate and mixed experiments on neural network clustering are also performed, based on fuzzy logic and simulated annealing algorithm, using Cauchy. Availability, necessary time and energy change of clustering are analyzed and concluded.(3) A new Fuzzy clustering algorithm NFC based on fuzzy logic neural network and simulated annealing algorithm is proposed, which use fuzzy logic neural network and Cauchy simulated annealing algorithm to find the cluster center, then clustering with FCM algorithm. Experiments show NFC is not sensitive to initial random center, and has good performance with a success rate of 99.76 percent, and avoiding the problem of local optimization.(4) NFC is applied to hydrology distribution of the Huaihe River in Anhui province. In which, the Main Factor Analysis is used to select relevant attribute, and NFC is used to clustering. The experiments on collected 124716 raw hydrologic data. Compared with traditional methods, it improves the time performance, accuracy and result reflects present state of hydrology distribution objectively. That is to help to plan the network of hydrology site and is convenient to be applied in the future.