基于参数自适应算法的环塔里木盆地红枣优生区划分  

Optimal-adaptive zone division of Chinese jujube in Tarim Basin based on Parameter Adaptive Algorithm

作　　者：金玉 徐焕良[1] 梁丰志 王江波[2] 王浩云[1] JIN Yu;XU Huanliang;LIANG Fengzhi;WANG Jiangbo;WANG Haoyun(Department of Computer Science,College of Artificial Intelligence,Nanjing Agricultural University,Nanjing 210095,Jiangsu,China;Faculty of Plant Sciences,Tarim University,Alar 843300,Xinjiang,China)

机构地区：[1]南京农业大学人工智能学院计算机系,南京210095 [2]塔里木大学植物科学学院,新疆阿拉尔843300

出　　处：《果树学报》2022年第1期95-103,共9页Journal of Fruit Science

基　　金：中央高校基本科研业务费专项(KYLH202006、KYZ201914);南京农业大学-塔里木大学科研合作联合基金(NNLH202006);新疆生产建设兵团南疆重点产业支撑计划(2017DB006);国家自然科学基金(31601545)。

摘　　要：【目的】进一步优化新疆红枣优生区,促进新疆红枣产业的提质增效。【方法】现有的优生区划分算法,由于其参数设值受人为主观影响较大,使优生区划分呈现较大的不确定性。以环塔里木盆地19个地区33个采样点的灰枣和骏枣的12个果实品质指标为数据集,提出了参数自适应的红枣优生区划分算法。【结果】首先对枣样本品质数据集进行300次有放回的随机抽样,对每一次抽取的样本进行参数自适应的主成分分析,并将分析结果进行融合得到红枣的主要品质指标。在此基础上,对各产区枣样本数据进行150次参数自适应的聚类分析,根据每次聚类结果所对应的红枣主要品质指标,利用无向加权图进行融合,得到不同品质指标所对应的优生区划分结果。确定灰枣和骏枣的主成分,在主要品质指标上将灰枣和骏枣优生区划分为4类。【结论】提出新疆环塔里木盆地各产区灰枣和骏枣的主要品质指标为总酸含量、总糖含量、单果质量和制干率,为新疆红枣区域化发展提供了依据。【Objective】Xinjiang has a unique advantage in light and heat resources,and the fruit in Xinjiang has very high quality.In response to the national policy of vigorously developing the jujube industry,under the goal of“improving the quality and efficiency”of the jujube industry in Xinjiang,the optimal-adaptive zone division should be taken more into account to meet the demand for the jujube quality.The existing algorithm of optimal-adaptive zoning presents great uncertainty because the parameter values are greatly influenced by human subjectivity.【Methods】Taking 12 fruit quality indexes of Huizao jujube and Junzao jujube from 33 sampling points in 19 regions around Tarim Basin as data set,a parameter adaptive algorithm was proposed to divide the optimal-adaptive zones of Chinese jujube.The parameter adaptive principal component analysis algorithm adopted the method of constant extraction and replacement to do m random sampling of n sample data sets.Due to the randomness,m data sets were different.This method can play the role in expanding the data set,avoiding the algorithm targeting only to fixed data sets.The experiment was set to select the principal components with a contribution rate of 50%-90%and a step size of 10%for each group of sampled data set.The ranking of the selected principal components’contribution rate was scored to obtain the principal components of group M and the scores of each group of principal components.Finally,the comprehensive score of the principal components of group M was calculated to obtain the comprehensive selection results of principal components.Parameter adaptive clustering analysis algorithm was carried out to select quality index of dimension reduction after the data sampling of p,a sample data set,the K values for each group of sampling data set for 3,4,5,clustering analysis and calculation of each group of data similarity index,each group of data clustering results with the ideas of the undirected weighted graph clustering,and clustering adjacency matrix.Each v

关 键 词：红枣 优生区 果实品质 参数自适应算法 

分 类 号：S665.1[农业科学—果树学]