| China is a country abundant in marine resources. The sustainable development of the national economy can not be maintained without the rational exploitation and protection of the marine. Driven by the development of agriculture and industry, as well as the population explosion in China, the problem of marine environment pollution is getting increasingly serious. Harmful Algal Blooms (HAB) breaking out frequently and expanding constantly in scale, have disturbed the ecological balance and endangered human life and property safety, which result in huge economic losses. To satisfy the strategic plan of national oceanic development and the urgent needs of social development, it is necessary to research and construct a reliable marine environment on-line monitoring system, and a precise and efficient HAB predicting system.First, by referring to the successful experience of ocean tri-dimensional monitoring system at home and abroad and by relying on the increasingly mature marine monitoring technology combined with advanced computer technology, network communications technology, control theory, data warehouse technology and embedded systems, this thesis deeply studies and proposes the basic framework of Marine Monitoring and Harmful Algal Blooms Predicting System, and designs an integration protocol to support the interaction between distributed and heterogeneous subsystems. Secondly, according to the complex mechanism of the occurrence of HAB and the characteristics of the marine environmental data, we plan and design Marine Environmental Data Warehouse by using the advanced database technology and the building model of data warehouse. Thus, the Marine Environmental Data Warehouse not only provides its users with rich and reliable data sources so that they can analyse and utilize the marine environmental data from different levels and different perspectives, but also provides powerful guarantee for relevant data mining work. Thirdly, this thesis introduces in detail the related theoretical knowledge of data mining and its application, with the focus on the clustering analysis, especially fuzzy c-means (FCM) clustering algorithm. In light of the deficiency of the traditional FCM clustering algorithm, SWFCM clustering algorithm is put forward based on similarity relation to have effectively improved the running speed and accuracy of clustering results. SWFCM is applied to the analysis of marine data, and has achieved good results in the HAB predicting. Finally, supported by various data mining algorithms, this thesis designes a visual Marine Environmental Data Mining System which is open-source, cross-platform and transplantable. The defects of complex data mining algorithms are remedied through intuitive visualization technology and the results of data mining are interpreted from the view-point of visibility, thus the functions of data product subsystem perfected.Marine Environmental Data Mining System is composed of two parts: Data Mining Software (MDMS) based on C/S mode and Web Data Mining System based on B/S mode. MDMS is designed under linux operation system by using MySQL and Qt Designer. The software is divided into four functional modules, start module, data acquisition module, data mining module and results display module. The modular structure of the software enhances its expansibility and configurability. It integrates a marine data receiving server, which can receive real-time marine data from monitoring subsystem and retransfer data to marine environmental data warehouse. MDMS can deal with various marine data sources, and its data mining module integrates various data mining algorithms. Combined with the characteristics of HAB, the algorithms are optimized, which effectively improves the practicability of the algorithm and the accuracy of prediction. In order to realize the Web Data Mining System based on B/S mode, LMAP is created. LMAP is a combination of open-source softwares, such as Linux, MySQL, Apache and PHP. Under the B/S mode, users send requests to server through the browser, and then the server executes users' requests and returns the results. Users c an achieve the functions such as query, analysis, and preservation of data simply throuth the browser. This mode greatly improves the system's usability, reduces the system's deployment and maintenance cost, and improves the visual degree of data mining.This subject is originated from Science and Technology Development Plan (Key Project) 2004GG2205108 of Shandong Province in 2004. The successful application of data mining technology to Marine Monitroing and Harmful Algal Blooms Predicting System contributes to the building, transforming and upgrading of the existing monitoring system and information platform, enhances the forecasting and prewarning capabilities of HAB, and protects the sustainable development of China's marine economy. |
PDF Full Text Request