Based On Fuzzy Clustering Decision Tree Germplasm In Maizestudy On Screening

Decision tree inductive algorithm is one of supervised machine learning algorithms. Itclassifies the dataset by building a tree model. Clustering algorithm is kind of unsupervisedmachine learning algorithms. It divide the dataset into lots of cluster by distance or someothers method. Nowadays, more and more scholar focus on combining the decision tree andclustering, divide the dataset into cluster with more accuracy by considering the all attributesof data.In this thesis, we study an algorithm named unsupervised fuzzy clustering decision tree.This method can divide the unlabelled dataset into cluster. This thesis presents a new nodeconsistency criterion, so promote the traditional fuzzy clustering decision tree algorithm intounsupervised learning area. Establishing the unsupervised decision tree is actually iterativeprocess of clustering algorithm. The fuzzy C-means clustering algorithm divides the dataset,then judge the tree structure by decision tree stability criterion. If the tree is stable, thenchoose the next splitting node by node consistency criterion. So, according to refining anddataset iteratively, a tree model is finally established.The experiment has two parts, in the first part, the experiment results show thatunsupervised fuzzy clustering decision trees can achieve better performance comparing totraditional clustering algorithm in several UCI datasets. In the second part, the proposedmethod is applied in corn germplasm chosen. By measuring the quality of corn seeds, it isconcluded that the two kinds of corn seeds can grow in the low potassium soil. Thisexperiment result also validate that the proposed method is effective in actually.