聚类支持下决策树模型在耕地评价中的应用

Application of evaluation in farmland with decision tree model based on clustering

为了挑选合理的学习样本,增强决策树模型在耕地评价应用的有效性,运用聚类方法挑选学习样本,用改进C 5.0决策树算法建立耕地评价模型,提出一种新的评价思路。运用此方法以广东省龙川县耕地为研究对象,以试验法挑选出6种聚类结果的学习样本,确定4000个样本作为最终的学习样本;利用决策代价权重来改进决策树评价模型,最终建立的评价模型的预测精度达到94.92%,满足了实际情况的需要。试验结果表明综合运用聚类和决策树模型进行耕地评价是可行的,其建立的评价模型具有精度高、鲁棒性和易理解性等特点。

To choose reasonable learning samples and to enhance the validity of decision tree in farmland evaluation, a new evaluation method Was proposed in the paper. The model of farmland evaluation was constructed by improved the algorithm of C5.0 decision tree based on clustering methods for choosing learning samples. Taking farmland in Longchuan County, Guangdong Province as example, six kinds of learning samples were chosen from clustering results by means of experimentation, and 4000 samples were finally learning samples. The evaluation model of decision tree was improved with cost weight, its finally prediction accuracy reached 94. 92%, which was satisfied with practical demand. The results show that the integrated mode using clustering method and decision tree is feasible for farmland evaluation. The constructed evaluation model has high accuracy, robustness and comprehension.