共查询到17条相似文献，搜索用时 62 毫秒
1.
随机森林模型在分类与回归分析中的应用 总被引：25，自引：0，他引：25
随机森林(random forest)模型是由Breiman和Cutler在2001年提出的一种基于分类树的算法。它通过对大量分类树的汇总提高了模型的预测精度,是取代神经网络等传统机器学习方法的新的模型。随机森林的运算速度很快,在处理大数据时表现优异。随机森林不需要顾虑一般回归分析面临的多元共线性的问题,不用做变量选择。现有的随机森林软件包给出了所有变量的重要性。另外,随机森林便于计算变量的非线性作用,而且可以体现变量间的交互作用(interaction)。它对离群值也不敏感。本文通过3个案例,分别介绍了随机森林在昆虫种类的判别分析、有无数据的分析(取代逻辑斯蒂回归)和回归分析上的应用。案例的数据格式和R语言代码可为研究随机森林在分类与回归分析中的应用提供参考。 相似文献
2.
森林类型的空间分布是进行森林景观格局研究的基础和先决条件。当前林业实际生产过程中 ,森林分布图的获得要么具有很强的主观性 ,要么受环境因素的制约。空间统计学可以描述事物在空间上的分布特征 ,条件模拟算法是空间统计学中进行空间插值的一种有效手段。森林类型是一个区域化的分类变量 ,在研究区内可能存在森林类型破碎化严重情况 ,为此本文选用分类变量的序列指示条件模拟算法来模拟森林类型的空间分布。文中介绍了序列指示条件模拟算法的原理、计算步骤、优点及适用性 ,以东北汪清林业局局级样地为材料 ,应用序列指示条件模拟算法对汪清林业局森林类型空间分布进行模拟 ,模拟结果与森林经理调查得到的森林分布图相比较 ,模拟精度达到 73.80 %。精度分析结果表明 ,以样地为材料 ,应用序列指示条件模拟算法 ,可以作为获得森林类型分布图的一个有效途径 相似文献
3.
比较3个应用较广的模拟物种地理分布模型:广义线性模型(GLM)、广义加法模型(GAM)与分类回归树(CART)对中国树种地理分布模拟的优劣,以提出更为合适的模拟物种地理分布模型,并用于预测气候变化对物种地理分布的影响。3个模型对中国15种树种地理分布的模拟研究表明:除对油松、辽东栎分布的模拟精度稍差外,对其余树种分布的模拟精度均较高,其中以GAM模型最好。结合地理信息系统(GIS),比较分析了这3个模型对青冈、木荷、红松和油松4种树种的地理分布模拟效果,结果亦表明:这3个模型均能很好模拟青冈和木荷的地理分布,而GLM模型对红松分布的模拟结果不太理想,3个模型对油松分布的模拟结果均不甚理想,其中以GLM模型最差。基于3个模型对未来气候变化下青冈与蒙古栎地理分布的预测表明:GLM模型与GAM模型对青冈分布的预测结果较为接近,青冈在未来气候变化情景下向西和向北扩展,而CART模型预测青冈在未来气候变化情景下除有向西、向北扩展趋势外,广东和广西南部的青冈分布区将消失;3个模型均预测蒙古栎在未来气候变化情景下向西扩展,扩展面积的大小为:模型的模拟面积>模型>模型。 相似文献
4.
物种分布模型是预测评估气候变化对物种分布影响的主要工具。为了降低物种分布模型在预测过程中的不确定性,近期有学者提出了采用组合预测的新方法,即采用多套建模数据、模型技术,模型参数,以及环境情景数据对物种分布进行预测,构成物种分布预测集合。但是,组合预测中各组分对变异的贡献还知之甚少,因此有必要把变异组分来源进行分割,以更有效地利用组合预测方法来降低模型预测中的不确定性。以油松为例,采用8个生态位模型,9套模型训练数据,3个GCM模型和一个SRES(A2)排放情景,模型分析了油松当前(1961-1990年)和未来气候条件下3个时间段(2010-2039年,2040-2069年,2070-2099年)的潜在分布。共计得到当前分布预测数据72套,未来每个时间段分布数据216套。采用开发的ClimateChina软件进行当前和未来气候数据的降尺度处理。采用Kappa、真实技巧统计方法(TSS)和接收机工作特征曲线下的面积(AUC)对模型预测能力进行评估。结果表明,随机森林(RF)、广义线性模型(GLM),广义加法模型(GAM)、多元自适应样条函数(MARS)以及助推法(GBM)预测效果较好,几乎不受建模数据之间差异的影响。混合判别分析模型(MDA)对建模数据之间的差异非常敏感,甚至出现建模失败现象。采用三因素方差分析方法对组合预测中的不确定性来源进行变异分割,结果表明,模型之间的差异对模拟预测结果不确定性的贡献最大且所占比例极高,而建模数据之间的差异贡献最小,GCM贡献居中。研究将有助于加深对物种分布模拟预测中不确定性的认识。 相似文献
5.
单木分割对于森林资源调查具有重要的意义,不同结构复杂度的森林单木分割算法的选择以及分割参数的选取对分割精度有着很大的影响。以山东田横岛为研究区,基于无人机正射影像与激光雷达数据,首先提取海岛森林典型植被二维与三维特征,然后利用随机森林算法对不同树种的树木进行分类,最后基于分类后的点云数据,选取不同结构复杂度的森林样地,对比分析聚类算法、层堆叠算法、分水岭算法在不同复杂度林区的适用性。结果表明：（1）随机森林算法结合单木二维、三维特征可有效对混交林树木进行分类,模型总体的精度为94.51%,Kappa系数为0.9038;（2）聚类算法对结构简单的林区具有更高的分割精度（F=96.41）,但依赖于分割参数的选取;面对复杂单木集群,分水岭算法总体得分波动最小（ΔF=14.56）,表现出较强的稳定性;（3）混交林预先进行树种分类可有效改善单木分割环境,相比于直接进行单木分割,聚类算法、层堆叠算法、分水岭算法的分割精度均得到不同程度的提升（ΔF1=10.06,ΔF2=9.51,ΔF3=12.6）。 相似文献
6.
明确野生动植物的地理分布是基础生态学和应用生态学领域的一个基础但关键的步骤,为后续分析提供了重要的信息。而野生动植物分布调查是一项需要投入大量人力,精力和资金的工作,特别是稀有物种的调查。物种分布模型越来越受到广泛引用尤其是在生物保护方面。为了证明物种分布模型在野生生物调查中精确采样方法的可行性,以全球易危物种黑颈鹤和白头鹤的实际繁殖分布预测为例,使用随机森林(Random Forest)算法加以验证。比较发现物种分布模型预测实际调查分布点,随机样方法生成的随机点,系统样方法的规则点在空间相对出现概率具有显著差异(P0.001),实际分布点具有较高的相对出现概率。该结果表明若在物种分布相对出现概率较高区域设置样方能够减少实际调查区域,有效提高发现目标物种的概率,从而减少调查投入。基于物种分布模型的精确采样方法将有效地提高我们对稀有物种分布的了解,有利于野生动植物的保护规划。 相似文献
7.
作为草地资源大国,我国正面临严峻的草场退化形势。掌握草场植被盖度的历史演变趋势,是草场退化驱动力识别及风险评估的基础。目前已有研究多以参数回归方法估算植被盖度,但并未充分考虑其苛刻的使用条件。利用Landsat系列卫星遥感影像及地面植被盖度监测资料建立非参数回归——随机森林回归模型,并与传统线性回归方法进行比较,在此基础上应用随机森林回归模型估算近10年来布尔津县草场植被盖度的变化趋势,并对结果的不确定性进行分析。结果显示:传统的线性回归方法很难满足其基本的统计学假设条件,而随机森林模型不但无需进行假设条件检验,而且预测的准确性也优于以往普遍应用的线性模型。基于Landsat ETM+标准数据得到的反演结果较之TM和OLI数据普遍偏小,地表反射率数据虽然可以大幅降低传感器不同对反演结果所造成的影响,但结果仍存在约±10%的不确定性。涉及的草场类型众多,为了提高反演精度,后续研究需要分别计算其植被指数,并尽量减低传感器差异带来的不确定性。 相似文献
8.
以多功能为目标的森林模拟优化系统(FSOS)的算法与应用前景 总被引：3，自引：0，他引：3
森林模拟优化模型(forest simulation and optimization system,FSOS)已在加拿大不列颠哥伦比亚省和中国长白山区得到广泛应用.FSOS模型基于多种资源协调平衡管理观点,采用金属模拟退火优化算法安排森林经营作业方案,在实现森林经营多目标长期可持续协调发展的基础上达到森林资源的理想状态.FSOS模型中的多功能(或多目标)包括水的储存和净化、二氧化碳的截获、野生动物生境保护、生物多样性维持、可视景观质量与木材生产等,其目标的参照系“森林理想状态”由专家、环境组织和政府政策根据生态系统的多功能进行综合界定.本文详细介绍了FSOS基本参数和理想状态的界定,及金属模拟退火优化算法在森林生态系统规划中的应用,为实现森林生态系统多种资源的规划管理以及政府定量分析和管理森林生态系统的多种资源并有效监督森林作业和多种资源变化、实现森林资源可持续利用提供帮助. 相似文献
9.
森林是生态系统的重要组成部分,准确估算森林碳储量及其分布对于评价森林生态系统的功能具有重要意义。以龙泉市为研究区,利用2009年99个森林资源清查样地数据和同年度Landsat TM影像数据,采用高斯序列协同仿真(SGCS)与BP神经网络方法(BPNN)分别模拟森林地上部分碳密度及其分布,并进行了对比分析。随机将样本数据分成70个建模样本和29个检验样本。通过模型检验,BP神经网络预测值与实测值的相关性达到0.67,相对均方根误差为0.63,空间仿真方法预测值与实测值的相关性为0.68,相对均方根误差为0.63,空间仿真方法预测能力略高于神经网络方法。仿真结果表明,基于BP神经网络模拟的森林碳总量为11042990 Mg,平均碳密度为36.10 Mg/hm2,总体森林碳密度均值高于样地平均值8.82%。基于空间仿真模拟的森林碳总量为11388657 Mg,平均碳密度为37.23 Mg/hm2,总体森林碳密度均值高于样地平均值9.40%。对比分析可知:高斯协同仿真模拟和BP神经网络虽然在碳总量估算值上与抽样数据估计值相近,但两种方法在估测值的频率分布以及研究区碳分布上有较大的差异。与BP神经网络相比,序列高斯协同模拟结果更接近系统抽样样地实测值,全部样地预测值与实测值的相关性达到0.75,在估计区域森林碳空间分布上有明显优势。在碳密度值域与频率分布方面,序列高斯协同模拟结果分布更合理。综上所述,序列高斯协同模拟在森林碳空间估计方面要优于BP神经网络。 相似文献
10.
新型统计方法和多源、多尺度空间信息数据的产生促进了物种空间分布模型的快速发展。不同的物种空间分布模型在生态学理论的运用以及前提假设上存在差异。选用不同的模型方法和输入数据会带来预测结果的不确定性。对比并集成多个物种空间分布模型,同时利用多组输入数据可降低预测的不确定性,提高物种分布模拟的精度。本文以中国特有种铁杉（Tsuga chinensis）为例,运用基于R语言开发的BioMod软件包对比9个物种空间分布模型对铁杉的模拟效果。最后以曲线下面积（ROC）为权重集成9个模型的模拟结果,产生和筛选最佳的铁杉潜在空间分布图。研究发现随机森林模型（RF）的模拟效果最好,其次是多元适应回归样条函数模型（MARS）和广义相加模型（GAM）,模拟效果最差的是表面分布区分室模型（SRE）。模型集成结果显示,最适宜铁杉分布的区域集中在中国的西南及四川盆地周围,其次零星分散于华南和台湾部分地区。这一结果与前人对铁杉自然分布的描述和研究结果较为吻合。研究进一步表明,通过模型的集成能有效地降低由于单个模型所带来的模拟结果不确定性,从而提高模拟的精度和效果。 相似文献
11.
Association classification of a 30 hm2 dynamics plot in the monsoon broad-leaved evergreen forest in Pu’er,Yunnan, China 
下载免费PDF全文
下载免费PDF全文 《植物生态学报》2020,44(3):236
季风常绿阔叶林是我国南亚热带典型的地带性植被, 建立森林动态监测样地是研究生物多样性维持和群落构建机制的重要平台。该文以普洱30 hm 2森林动态监测样地为研究对象, 采用多元回归树、重要值、主成分分析与指示种相结合的方法对样地内750个样方进行群丛数量分类, 以获取季风常绿阔叶林森林植被群丛类型。结果表明: 森林动态监测样地共发现木本植物271种, 隶属于78科178属, 群落类型为短刺锥(Castanopsis echidnocarpa)群系, 可划分4个群丛类型, 分别是珍珠花+毛银柴-短刺锥+泥柯群丛(Lyonia ovalifolia + Aporosa villosa - Castanopsis echidnocarpa + Lithocarpus fenestratus Association)、云南瘿椒树+耳叶柯-短刺锥+西南木荷群丛(Tapiscia yunnanensis + Lithocarpus grandifolius - Castanopsis echidnocarpa + Schima wallichii Association)、大果杜英+黄药大头茶-短刺锥+西南木荷群丛(Elaeocarpus sikkimensis + Polyspora chrysandra - Castanopsis echidnocarpa + Schima wallichii Association)、西桦+尼泊尔桤木-短刺锥+枹丝锥群丛(Betula alnoides + Alnus nepalensis - Castanopsis echidnocarpa + Castanopsis calathiformis Association)。群丛间物种分布存在较多过渡重叠, 指示物种是区分群丛类型的主要依据, 海拔与坡向对群丛分类有较大的影响, 坡度对群丛分类影响较小。 相似文献
12.
Aim Lepidium latifolium (Brassicaceae; perennial pepperweed) is a noxious Eurasian weed invading riparian and wetland areas of the western USA. Understanding which sites are most susceptible to invasion by L. latifolium will allow more efficient management of this weed. We assessed the ability of advanced remote sensing techniques to develop habitat suitability models for L. latifolium .
Location San Francisco Bay/Sacramento-San Joaquin River Delta, California, USA.
Methods Lepidium latifolium distribution was mapped with hyperspectral image data of Rush Ranch Open Space Preserve, providing presence/absence data to train and validate habitat models. A high-resolution light detection and ranging digital elevation model was used to derive predictor environmental variables (distance to channel, distance to upland, elevation, slope, aspect and convexity). Aggregate decision tree models were used to predict the potential distribution of this species.
Results Lepidium latifolium infested two zones: near the marshland–upland margin and along channels within the marsh. Topographical data, which are typically strongly correlated with wetland species distributions, were relatively unimportant to L. latifolium occurrence, although relevant microtopography information, particularly relative elevation, was subsumed in the distance to channel variable. The map of potential L. latifolium distribution reveals that Rush Ranch contains considerable habitat that it is susceptible to continued invasion.
Main conclusions Lepidium latifolium invades relatively less stressful sites along the inundation and salinity gradients. Advanced remote sensing datasets were shown to be sufficient for species distribution modelling. Remote sensing offers powerful tools that deserve wider use in ecological research and management. 相似文献
Location San Francisco Bay/Sacramento-San Joaquin River Delta, California, USA.
Methods Lepidium latifolium distribution was mapped with hyperspectral image data of Rush Ranch Open Space Preserve, providing presence/absence data to train and validate habitat models. A high-resolution light detection and ranging digital elevation model was used to derive predictor environmental variables (distance to channel, distance to upland, elevation, slope, aspect and convexity). Aggregate decision tree models were used to predict the potential distribution of this species.
Results Lepidium latifolium infested two zones: near the marshland–upland margin and along channels within the marsh. Topographical data, which are typically strongly correlated with wetland species distributions, were relatively unimportant to L. latifolium occurrence, although relevant microtopography information, particularly relative elevation, was subsumed in the distance to channel variable. The map of potential L. latifolium distribution reveals that Rush Ranch contains considerable habitat that it is susceptible to continued invasion.
Main conclusions Lepidium latifolium invades relatively less stressful sites along the inundation and salinity gradients. Advanced remote sensing datasets were shown to be sufficient for species distribution modelling. Remote sensing offers powerful tools that deserve wider use in ecological research and management. 相似文献
13.
Samuel Bosch Lennert Tyberghein Klaas Deneudt Francisco Hernandez Olivier De Clerck 《Diversity & distributions》2018,24(2):144-157
Aim
Ideally, datasets for species distribution modelling (SDM) contain evenly sampled records covering the entire distribution of the species, confirmed absences and auxiliary ecophysiological data allowing informed decisions on relevant predictors. Unfortunately, these criteria are rarely met for marine organisms for which distributions are too often only scantly characterized and absences generally not recorded. Here, we investigate predictor relevance as a function of modelling algorithms and settings for a global dataset of marine species.Location
Global marine.Methods
We selected well‐studied and identifiable species from all major marine taxonomic groups. Distribution records were compiled from public sources (e.g., OBIS, GBIF, Reef Life Survey) and linked to environmental data from Bio‐ORACLE and MARSPEC. Using this dataset, predictor relevance was analysed under different variations of modelling algorithms, numbers of predictor variables, cross‐validation strategies, sampling bias mitigation methods, evaluation methods and ranking methods. SDMs for all combinations of predictors from eight correlation groups were fitted and ranked, from which the top five predictors were selected as the most relevant.Results
We collected two million distribution records from 514 species across 18 phyla. Mean sea surface temperature and calcite are, respectively, the most relevant and irrelevant predictors. A less clear pattern was derived from the other predictors. The biggest differences in predictor relevance were induced by varying the number of predictors, the modelling algorithm and the sample selection bias correction. The distribution data and associated environmental data are made available through the R package marinespeed and at http://marinespeed.org .Main conclusions
While temperature is a relevant predictor of global marine species distributions, considerable variation in predictor relevance is linked to the SDM set‐up. We promote the usage of a standardized benchmark dataset (MarineSPEED) for methodological SDM studies.14.
Nocturnal hypoglycemia is a common phenomenon among patients with diabetes and can lead to a broad range of adverse events and complications. Identifying factors associated with hypoglycemia can improve glucose control and patient care. We propose a repeated measures random forest (RMRF) algorithm that can handle nonlinear relationships and interactions and the correlated responses from patients evaluated over several nights. Simulation results show that our proposed algorithm captures the informative variable more often than naïvely assuming independence. RMRF also outperforms standard random forest and extremely randomized trees algorithms. We demonstrate scenarios where RMRF attains greater prediction accuracy than generalized linear models. We apply the RMRF algorithm to analyze a diabetes study with 2524 nights from 127 patients with type 1 diabetes. We find that nocturnal hypoglycemia is associated with HbA1c, bedtime blood glucose (BG), insulin on board, time system activated, exercise intensity, and daytime hypoglycemia. The RMRF can accurately classify nights at high risk of nocturnal hypoglycemia. 相似文献
15.
林火是森林生态系统中重要的干扰因子之一,深刻地影响森林景观结构和功能。在全球气候化背景下,揭示气候变化对林火空间分布格局的影响,可为林火管理和防火资源分配提供科学指导。因此,基于江西省2001—2015年MODIS火影像数据(MCD14ML)和年均气温、年均降水量、植被、地形、人口密度、距道路距离、距居民点距离7个因子数据,利用增强回归树模型:(1)分析林火发生影响因子的相对重要性及其边际效应;(2)将GFDL-CM3和GISS-E2-R气候变化模式中的年均气温和年均降水量作为未来的气象数据,在3个温室气体排放量情景(RCP2.6、RCP4.5、RCP8.5)下,对2050年(2041—2060的平均值)和2070年(2061—2080的平均值)江西省林火分布进行预测,生成林火发生概率图。并采用受试者工作特征(ROC曲线)和混淆矩阵评估模型预测的精度。研究结果表明:(1)年均气温和海拔与江西省林火发生的相关性较强,年均降水量、居民点距离、人口密度、道路距离与林火发生的相关性较弱,但是与林火发生密切相关的如降水、风速等也应重点关注;(2)训练数据(70%)和验证数据(30%)的AUC值(ROC曲线下面积值)均为0.736,混淆矩阵对火点预测的正确率为67.8%,表明模型能够较好地预测研究区林火的发生;(3)在RCP8.5排放情景中林火发生的增幅最明显,其增幅较大的区域由赣南向赣北移动;(4)未来2050年和2070年林火发生与当前气候(2001—2015年)下相比,赣州市、鹰潭市的增幅较为明显,其他区域不明显。江西省各林业管理部门要加强林火高发区及潜在发生区的森林监测和管理,加大防火宣传力度,提升民众的森林防火意识。 相似文献
16.
Matthew S. Nolen Daniel D. Magoulick Robert J. DiStefano Emily M. Imhoff Brian K. Wagner 《Freshwater Biology》2014,59(11):2374-2389
- Crayfishes and other freshwater aquatic fauna are particularly at risk globally due to anthropogenic demand, manipulation and exploitation of freshwater resources and yet are often understudied. The Ozark faunal region of Missouri and Arkansas harbours a high level of aquatic biological diversity, especially in regard to endemic crayfishes. Three such endemics, Orconectes eupunctus, Orconectes marchandi and Cambarus hubbsi, are threatened by limited natural distribution and the invasions of Orconectes neglectus.
- We examined how natural and anthropogenic abiotic factors influence these three species across multiple spatial scales. Local and landscape environmental variables were used as predictors in classification and regression tree models at stream segment and segmentshed scales to determine their relation to presence/absence and density of the three species.
- Orconectes eupunctus presence was positively associated with stream size, current velocity and spring flow volume. Orconectes marchandi presence was predicted primarily by dolomite geology and water chemistry variables. Cambarus hubbsi was associated with larger stream size, with highest densities occurring in deep waters. Stream segment and segmentshed scale models were similar, but there were important differences based on species and response variables (presence/absence versus density). Stream segment scale models consistently performed better than or equal to segmentshed scale models.
- Anthropogenic abiotic environmental variables were of minor importance in most models, with the exception of O. marchandi being negatively related to road density and human population density. Classification tree models predicting distribution performed well when compared to random assignment, but regression trees were generally poor in explaining variation in density.
- We found that a range of environmental variables were important in predicting crayfish distribution and abundance at multiple spatial scales and their importance was species‐, response variable‐ and scale dependent. We would encourage others to examine the influence of spatial scale on species distribution and abundance patterns.
17.
Yong Cao Alison Stodola Sarah Douglass Diane Shasteen Kevin Cummings Ann Holtrop 《Freshwater Biology》2015,60(7):1379-1397
- Freshwater mussels are one of the most imperilled animal groups in the world. Their effective conservation and restoration require a better understanding of their spatial distributions at a relevant scale and of their relationships with natural environmental factors and human disturbances.
- In this study, we sampled over 900 sites on wadeable streams throughout Illinois, U.S.A., and compiled environmental data for a wide range of natural and anthropogenic factors related to climate, geology, land use, and connections to large rivers, dams and ponds.
- Using random forest classification and regression, we modelled the presence–absence of mussels as a group (87% accuracy), the abundances of 29 individual mussel species (R2 = 0.2–0.51), species richness (R2 = 0.52) and total mussel abundance in a standard sample (R2 = 0.41).
- The abundances of most species increased with stream size, the proportion of agricultural land in the catchment and the distance to the nearest dam or pond, but decreased with increasing catchment or channel slope and the proportion of forest in the catchment. Species varied in their relationships with climate variables, suggesting that they respond differently to climate change. Geology, particularly bedrock depth, was important for many species. Species richness and total mussel abundance responded positively to stream size and negatively to the slope of streams or catchments.
- The models were applied to unsampled wadeable stream reaches to generate mussel distribution maps at the reach scale, useful tools for resource managers to effectively protect and restore mussel biodiversity. The models also improve our understanding of how mussel populations and assemblages are structured by natural factors and human disturbances at a broad scale.