文献类型：期刊文献

中文题名：基于粗糙集与C5.0决策树的林地质量评价

英文题名：Evaluation of forest land quality based on rough set and C5.0 decision tree

作者：张宗艺[1,2] 刘鹏举[2] 唐小明[1,2]

第一作者：张宗艺

机构：[1]北京林业大学水土保持学院;[2]中国林业科学研究院资源信息研究所

年份：2017

卷号：45

期号：3

起止页码：96-102

中文期刊名：西北农林科技大学学报：自然科学版

收录：CSTPCD;;北大核心:【北大核心2014】；CSCD:【CSCD2017_2018】；

基金：中央级公益性科研院所基本科研业务费专项"基于大数据理论的森林资源数据空间分析技术研究"(IFRIT201502)

语种：中文

中文关键词：粗糙集;决策树;林地质量评价;森林资源小班数据;胡桃楸

外文关键词：rough sets; decision tree; forest land quality grade; forest resources subcompartment data; Juglans mandshurica

分类号：S757

摘要：【目的】使用数据挖掘算法实现多因子共同影响下的林地质量综合评价,探索林地质量与环境因子之间的非线性关系,为提高森林经营信息化水平提供技术支持。【方法】使用辽宁省抚顺市胡桃楸(Juglans mandshurica Maxim)森林资源小班数据,采用粗糙集算法筛选出与林地质量相关的重要因子,然后建立C5.0决策树,得出环境因子与林地质量间的非线性关系。【结果】影响胡桃楸林地质量的主要因子有坡度、坡向、坡位、海拔、下木种类、下木盖度、地被物种类、地被物盖度和土壤质地;以粗糙集方法选取的因子为输入变量的决策树模型规模小、复杂度低、决策规则简单,预测准确率达91.20%。【结论】本研究提出的林地质量等级预测和评价方法,能在保证模型准确率的同时降低算法的时间和空间复杂性,提高数据挖掘效率,并能克服一般林地质量评价中靠专家打分的局限性与主观性。
【Objective】This study used data mining algorithm to realize the comprehensive evaluation of forest land quality under the influence of multiple factors and explored the nonlinear relationship between forest land quality and environmental factors to provide technical support for improving information level of forest management.【Method】Based on the Juglans mandshurica Maxim forest resources subcompartment data in Fushun,Liaoning,the important factors related to the quality of forest land were selected by rough set algorithm,and the C5.0decision tree was established to obtain the nonlinear relationship between environmental factors and forest land quality.【Result】The main factors affecting the quality of J.mandshurica Maxim forest land included slope,slope aspect,slope position,elevation,shrub species,shrub coverage,litter types,litter cover and soil texture.The decision tree model based on rough set method had smaller scale,lower complexity,and simple decision with accuracy rate of 91.20%.【Conclusion】The established method can not only evaluate forest land quality precisely,but also reduce the complexity.It improves the efficiency of data mining,and overcomes the limitation and subjectivity of expert scoring in traditional evaluation of forest land quality.