随机森林算法在区域生态旅游适宜性评价中的应用研究  

Application of Random Forest Algorithm in Regional Ecotourism Suitability Assessment

作　　者：谭翠[1,2] 黄钦 杨波 李涛[1,2] 雷济华 TAN Cui;HUANG Qin;YANG Bo;LI Tao;LEI Jihua(School of Geographical Sciences,Hunan Normal University,Changsha 410081,China;Hunan Key Laboratory of Geospatial Big Data Mining and Application,Hunan Normal University,Changsha 410081,China)

机构地区：[1]湖南师范大学地理科学学院,长沙410081 [2]地理空间大数据挖掘与应用湖南省重点实验室,长沙410081

出　　处：《地球信息科学学报》2024年第2期318-331,共14页Journal of Geo-information Science

基　　金：国家自然科学基金项目(41171342);湖南省教育厅重点项目(17A127)。

摘　　要：生态旅游适宜性评价是评估生态旅游发展潜力、制定生态旅游规划和进行生态旅游开发的基础和重要参照。本文引入机器学习方法,从方法可行性、数据映射和预测实现等方面进行探讨,应用随机森林算法对湖南武陵山片区生态旅游适宜性展开实证研究。湖南武陵山片区旅游资源丰富,脱贫后亟需开展生态旅游来巩固拓展脱贫攻坚成果实现与乡村振兴有效衔接以及促进可持续发展。研究结果表明:(1)将机器学习算法引入到区域生态旅游适宜性评价领域作为一种新方法,可为之后改进生态旅游适宜性评价方法提供新思路与新方案;(2)随机森林算法可以有效应用在区域生态旅游适宜性评价方面,可作为适宜性评价研究的一种新方法,模型优化后的平均测试精度达86.49%,受试者工作特征曲线(ROC)与坐标围成的面积(AUC)达0.95,评价结果能够准确反映湖南武陵山片区生态旅游适宜性程度;(3)特征重要性排序结果显示土地利用类型影响最大,占比达到28.98%,人口密度、距景点距离和生物丰富度等因子的影响也较大,分别为16.34%、12.2%和10.65%,在进行生态旅游开发时应重点考虑这些因素;(4)生态旅游适宜性结果表明,高度适宜与适度适宜区占比高,研究区生态旅游开发潜力大。根据不同适宜性结果提出不同的开发方向:高度适宜区走保护性开发模式,打造体验-教育型生态旅游;适度适宜区走联合性开发模式,打造支撑型生态旅游;边际适宜区走限制性开发模式;不适宜区应当禁止开发。针对研究结果提出“两中心一带一板块”开发策略,可为武陵山片区进行生态旅游开发及巩固脱贫成果提供理论和技术指导。The ecotourism suitability assessment is the basis and a crucial reference for evaluating development potential,formulating plans,and implementing exploitation in ecotourism.In this study,we first analyze the feasibility of machine learning methods for modeling ecotourism suitability,and the Random Forest(RF)algorithm is selected for conducting an empirical study in the Wuling Mountain area in Hunan Province.In the study area,there are abundant tourism resources with an urgent need for ecotourism development,which can not only consolidate and expand the achievements of poverty alleviation,but also effectively connect with rural revitalization,thereby promoting sustainable development of tourism.The results show that:(1) Machine learning,as a new regional ecotourism suitability assessment approach,provides new insights and solutions for further improvement of suitability assessment;(2) The RF algorithm as a typical machine learning method can be effectively applied in the regional ecotourism suitability assessment.The optimized RF model achieves an average testing accuracy of 86.49%,with an area under the curve(AUC) of 0.95.These results also indicate the ecotourism suitability of the Wuling Mountain area in Hunan Province;(3) The ranking of feature importance reveals that land use type contributes most to the model,accounting for 28.98%,followed by other significant factors including population density(16.34%),distance from scenic spots(12.2%),and biological richness(10.65%).The above factors should be all considered in ecotourism development efforts;(4) The ecotourism suitability results show a high proportion of highly and moderately suitable areas,suggesting significant potential for ecotourism development in the study area.Based on the ecotourism suitability assessment,different development directions are proposed:A protective pattern and experiential education-oriented ecotourism are well-suited in highly suitable areas;a joint pattern and supportive ecotourism are appropriate for moderately suitable areas;a

关 键 词：生态旅游 随机森林 适宜性评价 武陵山片区 机器学习 多源空间数据 潜力评估 可持续发展 

分 类 号：X322[环境科学与工程—环境工程] F592.7[经济管理—旅游管理]