Study On The Method Of Cultivated Land Quality Evaluation Based On BP Artificial Neural Network And Support Vector Machine

The quality of cultivated land is determined by the characteristics of all influencing factors and the influence of each other.The objective and accurate evaluation of the cultivated land quality has a guiding effect on agricultural production.Compared with the traditional method,the algorithm of the Artificial Neural Network(ANN)and the Support Vector Machine(SVM)do not need to get the weight artificially,it can automatically simulate the nonlinear relationship between the factors.In this paper,the hilly area of Changtai County in Fujian Province is used as an empirical area.Using MATLAB software as the operating platform,there will be a combination of supervised network learning(BP-ANN and SVM)and unsupervised network learning(SOM).First,selecting typical samples by using self-organizing feature mapping network(SOM).Then,7 indicators(effective soil thickness,organic matter content,slope,field road accessibility,altitude,soil texture and irrigation guarantee rate)were used as input variables,natural grade indexes or grade were used as output variables.Based on training and studying of BP-ANN and SVM,the BP-ANN classification model classification model and SVM classification model were established.The results show that the two models' accuracy met the expectations,but the SVM model is better than BP neural network and is more suitable for the evaluation of cultivated land quality.The study provided new techniques for the classification of cultivated land quality.(1)Using the BP-ANN as the evaluation technology,the data matrix(10225 × 7)is normalized to the value of 0 to 1,and the 2602 data sets are selected by the method of SOM clustering as training samples.Building network topology(7:17:1),and then training the network.The correlation coefficient is 0.988,the posterior difference ratio(C)is 0.16,the small error frequency(P)is 0.99,the accuracy grade is high.So,the BP neural network model can be used to calculate the other plots of cultivated land.Compared with the evaluation results of the application factor method,the correct rate of 4 grade is 94%,5 grade is 87%,6 grade is 60%,9 grade is 100%.The correct rate of final evaluation result is 89%.(2)Taking the SVM as the method of cultivating land quality evaluation,the samples are input into the SVM classifier,and the SVM model with different kernel functions is compared after running the learner.The SVM model is selected when the accuracy is up to 98.9%.Compared the results obtained by the SVM with the results obtained by the traditional method.The evaluation results are highly consistent with the actual values.So,this model can be applied to other plots of cultivated land.Compared with the evaluation results of the application factor method,the correct rate of 4 grade is 99%,5 grade is 98.5%,6 grade is 94%,9 grade is 100%.The correct rate of final evaluation result is 99%.(3)A comparative analysis of the BP-ANN and the SVM model.BP-ANN and SVM applied to the evaluation of cultivated land quality,they were strong learning ability.The application of 3-layer BP-ANN has the advantages of strong nonlinear transformation ability,large-scale data processing,self-learning and adaptive function.It runs quickly and precisely.It has a strong popularization function and robustness.SVM transforms the sample into a high-dimensional feature space under the nonlinear transformation of Cubic kernel function,and explores the distinguished support vector automatically.The classifier can realize the classification of unknown sample class.It also has a strong robustness and popularization.It does not need to adjust artificially.Its calculation is not only simple,but also accurate.However,the number of hidden nodes in the BP-ANN needs to be tested manually to obtain the optimal parameters,and the topology is difficult to determine.The SVM model can achieve the optimal parameters automatically in the construction,which can reduce the influence of human intervention.SVM convergence rate is faster in the evaluation process.The accuracy and popularization function of SVM model is higher than that of BP neural network model.In summary,the SVM is a better method for evaluating the quality of arable land.