Research And Application Of Pest Prediction Based On Rough Set And Artificial Neural Network

Artificial Neural Networks and Rough Sets are by uncertainty, ambiguity andincomplete treatment of the true value to find the best solutions, To dealing with theseissues, the two have separate systems approach, which obviously have complementaritiestechnical features . Therefore, the implementation of the integration of the two can bettersolve the problems.This paper focused on rough sets and artificial neural network, carried out theadvantages of the combination of research and discussion. Based on a combination of boththe relevant theoretical knowledge, the rough set attribute reduction algorithm, and rulesthrough the momentum factor combination of the improved BP algorithm, in this based onthe design and establishment of pest forecasting model, using the same data through theattribute of simple model with BP algorithm and has not been improved attribute reductionalgorithm with the improved BP model to compare, by test results showed that the modelbased on the improved model training time and prediction accuracy are improved on.Details are as follows:(1) On this research background, purpose and the significance, summarizing therecent artificial neural network and rough sets, respectively, in the prediction model oftheoretical research, on the basis of the rough set and neural network combinationResearch.(2) Introduce the background of this article in detailIncluding the basic principles of artificial neural networks and related concepts, suchas the classical model, also introduced the theory of rough sets and details of the rough setreduction algorithms for data processing and its research situation.(3) Analysis of the shortcomings of the standard BP and the lack of improved BPalgorithm and given the processRepeated in the neural network learning process, in the difference between networkoutput and actual output is large, the network global error rate of decline slowed down, noteven in decline, because the network convergence to local minimum, in order to avoid thissituation, We have revised the steepest descent method when the weight is not only amoment to consider the gradient, taking a moment before the moment of the gradient, sothat reduces the shock of learning process, thereby inhibiting the local minimum areanetwork, effective the neural network to avoid convergence to local minima problems.(4) The design is based on rough sets and artificial neural network prediction model ofpestUsing rough set attribute reduction algorithm processing the sample data set, remove redundant attributes from the sample as a neural network with input neurons as the outputlevel of pest outbreaks, the use of momentum factor improved by BP algorithm to train theneural network, came to the conclusion based on rough set and neural network predictionmodel combining pest.(5) Pest prediction models to achieve system and verified by experiments theadvantages of the improved modelWith the Matlab toolbox provided training to achieve the BP learning process, Matlabsimulation experiments, and with no standards for BP after data preprocessing andprediction models combining comparative analysis, the results show that the improvedmodel training time than the original model schools on the prediction accuracy rate and hasgreatly improved.This study is a rough set and artificial neural network model combining anoptimization in the handling uncertainty, ambiguity, the use of rough set neural networkinput on the pretreatment, not only reduces the complexity of the neural network, andimprove the prediction of neural network efficiency, while using the additional momentumand adaptive parameter adjustment method improved the traditional BP algorithm, furtherimproves the ability of rough neural application of the model, this study may be artificialintelligence, neural network applications, etc. have some theoretical significance andapplication value.