Corn Precision Fertilizer Decision System Based On Neural Network

Neural network ensemble has become a very hot topic of many academiccircles, artificial neural network is mathematical models used to simply simulate thehumanbrainmechanism,neuralnetworkadjust the internalconnectionsrelationshipbetween nodes according to the complexity of neural networks.This model in manyareas has been get many good results,,based on neural network model-Hlnenalgorithm for precision fertilization of corn decision-making system is a bettersolution to the inefficiency of traditional fertilizers, improving economic efficiencyand reduce pollution of the environment to achieve in precision agriculture andother innovative breakthrough in the design of a new type of neural networkensemble method.Although there are a number of neural network methods has agreat improvement compared to the traditionalstrategyof precisionfertilizereffects,but still revealed a number of shortcomings, such as the results of neural networkensemble is heavily dependent on the user's experience and proficiencybesides,the theoretical system is not precise enough, the algorithm still need to beimproved and so on.The neural network ensemble method is proposed of such shortcomingswhich not onlysolves the lackof points above, but also can increase the complexityof the generalization ability of neural networks, and improve the system accuracy,scholars have their eyes toward this field.Therefore, to achieve strong and effectiveneural network ensemble model is the great progress in which push the technologyinto the field of engineering.Currently, the main difficulties of the work in many decision-making field is howto improve forecasting and classification accuracy, no doubt,,integrating learningadvanced a feasible method for the above issues.Neural network ensemble modelis a effective way of integrating the field of forecasting and classification ,whichthrough improving the generalization ability of neural networkmethod ,thereby,,improve the computation accuracy.Precision agriculture isgradually grew up based on such merits, the labor productivity of traditional agricultural technology is lower,a lot of labor wasted on inefficient agriculturalproduction.In addition, there have been as soil erosion and emit large quantities ofpesticides led to environmental pollution, combined with the unity of genes lead todecreased agricultural production and many other problems.Under this pressure,scholars have proposed some programs .They find the new program about makefull use of the limited resources of the agricultural production which maintainagricultural production while also protecting the natural environment.Neural network ensemble method is proposed under such demand,neuralnetwork ensemble is a learning algorithm,which can construct a set of classificationand prediction, and organic synthesis the results.Neural network technology integration mainly referenced Bagging algorithms andBoostingalgorithms.EachneuralnetworktrainingsetofBaggingalgorithmfairlytheinitial training set of similar size, The selection of these examples can be repeated,so that the occurrence probability of each sample is different, the algorithm is anexample of the differences by increasing the degree of integration of to improveneuralnetworks generalization.TheBaggingalgorithmnotonlycansignificantlyimprovethegeneralizationabilityofthe system, but also can effectively simplify the calculation, so the learning hasbecome a research hotspot.Boosting algorithm is largely determined by the training set that arose prior tothe performance of neural networks, newneural network can be verydifficult for theneural network of treatment, therefore, the algorithm sometimes play a goodrole..Theoretical analysis, the design a more excellent neural network ensemble isverydifficult.The results show that the more differences of the training set between theindividual network, thesmaller generalization error and integration the better.Therefore, we should find ways to train great differences between individualnetworks.Usually,There are three categories in individual network.Training of dataperturbation methods, changing the characteristics of neural network methods,choose the method of integration in neural networks, and so on.In addition, thefindings generated combination of methods is also important because the methodcanbeintegratedinthevariousoutputdataforsimpleaverageorweightedaverage This paper is through a large number of theoretical analysis and experimentalexploration, identified a very suitable model for integrated neural network,Mixedlinear and nonlinear neural network ensemble method-Hlnen algorithm.Thecompletion of the work asfollows:First,In order to increase integration between the individual network degree,make full use of original data, as well as to make the original data set and thetraining data set similar.The different individual neural networks trained on differenttraining samples, can increase the differences between the various neural networklevel, in this regard to improve the generalization ability of neural networkensembles.Second, the work to be done is to use our neural network model canreduce the error, in order to solve the problem, the nitrogen, The two places of thedata was tested ,which contains phosphorus, potassium and other elements, andthe two land was tested with Hlnen algorithm.The results show a mixed linear and nonlinear neural network ensemblemethodproducestheerrorissmallerthanothermethods.Experimentalresultsshowthat the effectiveness of neural network,which the differences between the neuralnetwork generated through the integration.This means that the error of theirproduce in the different data space, the error between the various neural networkcan compensate each other, the effect can be more effectivelyintegrated.Probability of individuals from the perspective of the relationship between thenetwork and integration, the diversity of individual networks and neural networkintegration between the accuracy is not a simple linear relationship, this can beseen the core of this thesis.Currently ,the field is still in development stage , on thisissue ,we give the future direction and focusto work.This paper, the neural network ensemble status of domestic and foreignresearch Statementsthrough theoreticalanalysis, the method and the application ofthe results achieved, while also Referenced Formula, Main analysis for therealization of neural network ensemble method, and design a new type of neuralnetworkbased on improved integration of decision-makingsystem of corn precisionfertilizer.Thispaper isa simulation of thefarmland inYushu of Jilin Province, useingof the specificsoilsample data, carriedout soiltestingandfertilization, with thefinalproduction estimates,obtained the necessarydata sets.Experimental results show that the error of a mixed linear and nonlinear neuralnetwork ensemble method-Hlnen algorithm is less than linear and nonlinear integration ,more than a single neural network.This work has opened up a feasibleand effective path for artificial neural networks in practical applications inagriculture .