小波-智能优化支持向量机模型及其对农业可持续发展的评价

Wavelet-intelligent optimization model using support vector machine and its application for evaluating sustainable agricultural development

农业可持续发展评价是世界性农业研究的热点问题,也是中国新时代乡村振兴和农业现代化发展的重大课题。改革开放促进中国农业经济获得快速发展,但农业发展与农业资源及生态环境的矛盾严重。党的十八大以来,农业可持续发展被提到更加突出的位置。研究农业可持续发展评价具有重要意义。该文研究国内外农业可持续评价的相关文献,在小波变换、支持向量机、遗传算法、蚁群算法基础上,提出小波-智能优化支持向量机相结合评价方法,参考已有评价指标体系,结合中国的农业发展现实状况,建立农业可持续发展评价备选指标集。运用Spss软件对150个初选指标进行显著性和相关性分析,确立包含62个指标的评价指标体系。运用一维离散平稳小波分析,数据消噪处理,遗传算法(genetic algorithm,GA)、蚁群算法(ant colony optimization,ACO),优化支持向量机参数(support vector machine,SVM),得出较好的惩罚参数、核函数、不敏感系数,再对支持向量机训练,该方法提高了训练准确度。对中国31个省(市)农业可持续发展进行小波-遗传算法优化支持向量机(GA-SVM)、小波-蚁群算法优化支持向量机(ACO-SVM)评价,并与GA-SVM、ACO-SVM农业可持续评价进行比较。评价与仿真结果表明,小波-GASVM农业可持续发展评价均方误差为9.641×10^(-5),相关系数为0.980;而GASVM在同样的训练集测试集下,得到的均方误差、相关系数分别为0.006、0.979。小波-ACOSVM农业可持续发展评价均方误差为9.318×10^(-5),相关系数为0.972;而ACO-SVM在同样的训练集测试集下,均方误差MSE、相关系数分别为0.016、0.953,小波-GASVM、小波-ACOSVM的农业可持续发展评价在预测精度和收敛速度上分别优于GA-SVM、ACO-SVM两种方法,小波-GASVM的农业可持续评价效果更理想。通过改变测试集个数进行多次试验,同样验证小波-GASVM方法是可靠的。该研究为农业可持续发展评价提供方法参考。

Assessing the sustainable development of agriculture is a hot topic in global agricultural research and an important task of rural revitalization and agricultural upgrading in the new era.Reform and opening up promote the rapid development of the agricultural economy,but the conflict between agricultural development and agricultural resources and ecological environment is serious.Since the 18th National Congress of the CPC,the sustainable development of agriculture has been given more prominence.It is of great significance to study the evaluation of agricultural sustainable development.Firstly,this paper studies the relevant literature on agricultural sustainable development evaluation at home and abroad.Based on wavelet transform,support vector machine(SVM),genetic algorithm(GA)and ant colony optimization(ACO)algorithm,a combination evaluation method of wavelet and intelligent optimization support vector machine is proposed.Secondly,by referring to the existing evaluation index system and combining with the actual situation of China's agricultural development,the alternative index set of agricultural sustainable development evaluation is established.Using SPSS software,the significance and correlation of 150 primary indexes are analyzed,and an evaluation index system including 62 indexes is established.Thirdly,using one-dimensional discrete stationary wavelet analysis,data denoising processing,genetic algorithm,ant colony algorithm,optimize the parameters of support vector machine,get better penalty parameters,kernel function,insensitivity coefficient,and then train the support vector machine.Fourth,taking 31 provinces(municipalities)in China as an example,this paper evaluates the agricultural sustainable development by applying the wavelet-GASVM model and the wavelet-ACOSVM model,and using the GA-SVM model and the ACO-SVM model to evaluate the agricultural sustainable development of 31 provinces(municipalities)in China.Fifthly,the reliability of the wavelet intelligent optimization SVM method is verified by changing the number of test sets and conducting multiple tests.The evaluation and simulation results show that:1)The mean square error of agricultural sustainable development evaluation of Wavelte-GASVM is 9.641×10^(-5),and the correlation coefficient is 0.980;however,under the same training set and test set,the mean square error and correlation coefficient obtained by GASVM are 0.006 and 0.979,respectively.2)The mean square error of Wavelet-ACOSVM agricultural sustainable development evaluation was 9.318×10^(-5),and the correlation coefficient was 0.972;the mean square error MSE and correlation coefficient of ACO-SVM are 0.016 and 0.953,respectively,under the same training set and test set.3)Change the number of test sets and perform several tests and compare the results of GASVM model and ACO-SVM model before and after wavelet denoising respectively.It can be seen that under the GASVM model,the mean square error of the running results after wavelet denoising is smaller than that before wavelet denoising,and the correlation coefficient is larger than that before wavelet denoising,and the running time is shorter.Under the ACOSVM model,the mean square error of the running result after wavelet denoising is also smaller than that before wavelet denoising,and the correlation coefficient is larger than that before wavelet denoising,and the running time is shorter.Conclusion:The prediction accuracy and convergence speed of agricultural sustainable development evaluation by wavelet-GASVM and wavelet-ACOSVM are better than GA-SVM and ACO-SVM,respectively,and the effect of agricultural sustainable development evaluation by wavelet-GASVM is more ideal.This study provides a method reference for agricultural sustainable development assessment.