基于聚类的决策树在玉米种质筛选中的应用

Application of Clustering-based Decision Tree in the Screening of Maize Germplasm

[目的]建立一种改进的基于聚类的模糊决策树,并研究其在玉米种质筛选中的应用。[方法]采用一种新型的基于聚类的决策树算法,该算法针对传统的决策树算法不能处理无类别样本的这一不足,进行了改进。同时,将改进算法应用在玉米品种的筛选问题中,通过对叶面积、株高、干重、钾利用率等指标的衡量,筛选出耐低钾性较强的玉米种子。[结果]该算法在玉米种质的筛选上,适用性强且性能较优。[结论]在今后工作中还需进一步验证比较改进的基于聚类的模糊决策树与传统的模糊聚类决策树的性能,并将其应用在更多的实际问题中。

[Purpose] This paper aims to construct an improved fuzzy decision tree which is based on clustering,and researches into its application in the screening of maize germplasm.[Method] A new decision tree algorithm based upon clustering is adopted in this paper,which is improved against the defect that traditional decision tree algorithm fails to handle samples of no classes.Meanwhile,the improved algorithm is also applied to the screening of maize varieties.Through the indices as leaf area,plant height,dry weight,potassium（K） utilization and others,maize seeds with strong tolerance of hypokalemic are filtered out.[Result] The algorithm in the screening of maize germplasm has great applicability and good performance.[Conclusion] In the future more efforts should be made to compare improved the performance of fuzzy decision tree based upon clustering with the performance of traditional fuzzy one,and it should be applied into more realistic problems.