Design And Implementation Of Water Quality Index Prediction Model Based On BP Neural Network

As the resource for the survival of human beings,water should be strictly protected.In order to protect water resources and improve the increasingly serious water environment,experts and scholars proposed a variety of solutions and measures.This paper analysis the changing trends of water quality index and makes some corresponding solutions before situation get worse.The article presents a novel prediction method combined of neural networks and cross-validation method,the method used the optimized neural network for time series prediction.Based on the prediction model,we build the prediction platform and collect water quality index data.Using time series similarity data mining techniques to find monitoring points with similar characteristics,then using the actual water quality index of the similar monitoring point to predict the change trends of target point.The major task of this paper included:1.To understand the main factors affecting the water environment and water quality by consulting relevant literatures.Analyze and compare the advantages and disadvantages of each method to predict water quality index and mainly focused on using BP neural network.Given some optimization method include variable learning rate and adding momentum.2.Present the new prediction method that combined of an optimized neural network and cross-validation method,which is based on the research of the character of water quality index and neural network.The new forecasting methods are introduced and analyzed in detail,and the new prediction method was verified by experiment.Experimental result shows that the new combination forecasting solution can improve the stability and accuracy of the neural network prediction of water quality index.3.Analysis the similarity search techniques for time series,compared different representations of time series method,proposed one that based on the slope change representation method of time series.Experiment results show that this method is more suitable for the water quality index,because of the data size of the water quality index time series is smaller and it’s more sensitive to the local change of time series.4.Build a completely open platform for water quality index prediction based on the prediction model mentioned above.While user using the platform to predict the water quality index,platform collecting water quality index data thorough out the whole country.With the accumulation of data,we are able to analyze the correlation of each detection point.Then find out monitoring points with similar characteristics,by referring to actual data of similar points to predict the water quality index of target point.Finally,the article summarizes the technical and optimization method raised above,point out the shortcomings and propose possible optimization directions and objectives in the future work.