Research On Water Quality Data Mining And Application Based On Genetic Algorithm

With the rapid development of China's economy and population growth in coastal areas,the marine resources is excessively developed and improperly utilized,and the water pollution and ecological destruction in the offshore marine areas is growing worse and worse.To solve the problem of marine ecosystem effectively and protect the healthy and orderly development of marine ecosystem,it is necessary to provide a real time dynamic monitoring of the marine environment.Vast amounts of monitoring data can be acquired through monitoring platform.The traditional mathematical statistics and empirical prediction method are based on the experts' experience,in which the monitoring data have not been used fully.Therefore,it is a major problem that we should use advanced algorithms to get useful information through better use of the monitoring data,accurately assess the quality of marine water and establish an effective analysis and early warning model of marine water quality.This research relies on the water quality data monitored by the photoelectric sensor networks of the special optical fiber and fiber optic sensing key laboratory,analyzes the water quality monitoring data theoretically based on the summary of the current status of technology and research in water quality assessment,rule update and early warning model water quality analysis.Simulation results show the effectiveness of the algorithm,which is applied to the analysis of water quality monitoring data and the early warning.The main research contents include the following:Firstly,for the problem that massive water quality monitoring data has not been fully utilized,an association rule mining method based on adaptive immune genetic algorithm is researched.Immune algorithm and classical genetic algorithm are combined.Design fitness function.Crossover and mutation probability in immune genetic algorithm is improved so that it can be continuously changed along with individual fitness throughout the implementation process.And the algorithm is applied to mining association rules.The mining method in the rules can ensure accuracy,at the same time,can greatly shorten the time of excavation.Secondly,to solve the problem that water quality monitoring database is dynamically changing over time,and the accumulation of new data and the dynamic changes causes the maintenance problems,a rule update mining algorithm is researched based on the adaptive immune genetic association rule mining algorithm.With the update mining of property database,a new database is composed of part of the original data and the new data.Calculate the degree of difference of the original data and the new,then extract a certain percentage of data from the original database to form a new database,and mine the new database with the immune genetic association rules.The algorithm can not only retain the original high support rules,but also find new rules.Thirdly,for the problem of changing parameter,a BP neural network optimization method based on immune hierarchical genetic algorithm is researched in order to make the changing parameter applicable to the analysis of early warning model based on BP neural network.Water quality monitoring parameters can be changed in the actual testing process,and changing parameters would affect the structure of BP neural network.To solve the problem,the following measures should be taken: improve the algorithm processing speed by introducing immune algorithm;design the fitness function;using the chromosome stratification characteristics of the hierarchical genetic algorithm,parameters can be classified as parameter gene and control gene,then apply the decoding operation in the traditional genetic algorithms,which can optimize both structure and weights of neural network;adjust the crossover and mutation probability so that it can be continuously changed along with individual fitness throughout the implementation process.The method can adjust BP neural network structure according to dynamic changes in water quality monitoring parameters and enable the analysis warning model to be more applicable to the current parameters study and analysis and make early warning and timely.Finally,to improve the accuracy of water quality monitoring analysis and early warning model,the optimized BP neural network algorithm based on adaptive immune hierarchical genetic algorithm and Bayesian regularization LM-BP neural network are combined together and a new method to establish evaluation-warning model of water quality monitoring data is researched.The parameters and structure of the neural network can be optimized through immune hierarchical genetic algorithm.The accuracy of the evaluation-warning model can be improved through Bayesian regularization LM-BP neural network.This model is applicable to multi-parameter large sample training.