深度随机森林在离网预测中的应用

在电信运营商领域,离网预测模型是企业决策者用来发现潜在离网用户(即停用运营商服务)的主要手段。目前离网预测模型都是基于逻辑回归、决策树、神经网络及随机森林等浅层机器学习算法,但是在大数据的背景下,这些浅层算法在预测问题上很难取得更高的精度。因此,提出了一种新型的深层结构模型——深度随机森林,通过将传统浅层随机森林堆积成深层结构模型,获得更高的预测精度。在运营商真实数据上进行了大量实验,结果证明深层随机森林模型比传统浅层机器学习算法在离网预测问题上可以得到更好的效果。同时,增大训练数据量可以进一步提升深层随机森林的预测能力,从而证明了在大数据环境下深层模型的潜力。     

基于颜色和纹理特征的林火烟雾识别

为了实现森林火灾的智能化预警,提出了基于颜色和纹理特征的林火烟雾识别方法.首先使用颜色特征确定烟雾疑似区域,随后采用局部二值模式方差(Local Binary Pattern Variance,LBPV)提取疑似区域纹理的不规则度特征并产生LBP图像.然后利用小波变换分解LBP图像并提取模糊度、复杂度和相关度特征.最后利用支持向量机(Support Vector Machine,SVM)进行烟雾识别.结果证明,颜色结合纹理特征方法可以有效实现林火烟雾的识别,为林火烟雾识别研究提供了一种有效方案.     

一种改进的SVM算法在乳腺癌诊断方面的应用

针对计算机辅助诊断(CAD)技术在乳腺癌疾病诊断准确率的优化问题,提出了一种基于随机森林模型下Gini指标特征加权的支持向量机方法(RFG-SVM)。该方法利用了随机森林模型下的Gini指数衡量各个特征对分类结果的重要性,构造具有加权特征向量核函数的支持向量机,并在乳腺癌疾病诊断方面加以应用。经理论分析和实验数据验证,相比于传统的支持向量机(SVM),该方法提升了分类预测的性能,其结果与最新的方法相比也具有一定的竞争力,而且在医疗诊断应用方面更具优势。     

基于CMYK空间的火灾区域粒子群优化快速定位方法

将CMYK彩色模型下的火焰颜色特征与粒子群优化(Particle swarm optimization,PSO)算法求最优适应度函数相结合,通过合理配置检测窗口邻域中颜色的M,y各低阶矩的特征在欧氏距离分类适应度函数中的因子及PSO中各参数,使森林监测图像的可疑火灾着火点的初定位控制在0.05s以内,最后通过对火灾样本关键图与初定位可疑着火点的颜色互信息的计算而得到进一步的校验和确定.实验表明改进的PSO算法具有良好的火焰区域定位效果及运算效率,易于移植到前端探测的嵌入式火灾探测系统.     

结合Rotation Forest和MultiBoost的SVM集成方法

针对如何提高集成学习的性能,提出一种结合Rotation Forest和Multil3oost的集成学习方法—利用Rotation Forest中旋转变换的思想对原始数据集进行变换,旨在增加分类器间的差异度;利用Mu1tiI3oost在变换后的数据集上训练基分类器,旨在提高基分类器的准确度。最后用简单的多数投票法融合各基分类器的决策结果,将其作为集成分类器的输出。为了验证该方法的有效性,在公共数据集UCI上进行了实验,结果显示,该方法可获得较高的分类精度。     

Google Earth在森林植被营养元素循环研究中的应用

森林植被营养元素循环是森林生态系统分析中的重要参数.将Google Earth应用在其研究过程中具有重要作用.阐述了Google Earth的主要原理和关键技术,重点分析了KML和Google EarthAPI;设计了Google Earth应用的数据处理流程和系统架构;最后实现了基于Google Earth的贵州龙里森林植被营养元素循环研究系统.     

Active learning paradigms for CBIR systems based on optimum-path forest classification

This paper discusses methods for content-based image retrieval (CBIR) systems based on relevance feedback according to two active learning paradigms, named greedy and planned. In greedy methods, the system aims to return the most relevant images for a query at each iteration. In planned methods, the most informative images are returned during a few iterations and the most relevant ones are only presented afterward. In the past, we proposed a greedy approach based on optimum-path forest classification (OPF) and demonstrated its gain in effectiveness with respect to a planned method based on support-vector machines and another greedy approach based on multi-point query. In this work, we introduce a planned approach based on the OPF classifier and demonstrate its gain in effectiveness over all methods above using more image databases. In our tests, the most informative images are better obtained from images that are classified as relevant, which differs from the original definition. The results also indicate that both OPF-based methods require less user involvement (efficiency) to satisfy the user's expectation (effectiveness), and provide interactive response times.     

Large area forest stem volume mapping in the boreal zone using synergy of ERS-1/2 tandem coherence and MODIS vegetation continuous fields

ERS-1/2 tandem coherence was reported to have high potential for the mapping of boreal forest stem volume (e.g. Santoro et al., 2002, 2007a; Wagner et al., 2003; Askne & Santoro, 2005). Large-scale application of the data for forest stem volume mapping, however, is hindered by the variability of coherence with meteorological and environmental acquisition conditions. The traditional way of stem volume retrieval is based on the training of models, relating coherence to stem volume, with the aid of forest inventory data which is generally available for a few small test sites but not for large areas. In this paper a new approach is presented that allows model training using the MODIS Vegetation Continuous Fields canopy cover product (Hansen et al., 2003) without further need for ground data. A comparison of the new approach with the traditional regression-based and ground-data dependent model training is presented in this paper for a multi-seasonal ERS-1/2 tandem dataset covering several well known Central Siberian forest sites. As a test scenario for large-area application, the approach was applied to a multi-seasonal ERS-1/2 tandem dataset of 223 ERS-1 and ERS-2 image pairs covering Northeast China (~ 1.5 million km²) to map four stem volume classes (0-20, 20-50, 50-80, and > 80 m³/ha).     

林地调查中的面积单元线性测量法（SL法）

林业调查中采用的基本方法是全林调查法和局部估测法,而局部估测方法中最常用的是角规测量法与标准地法,在林地调查中角规法误差较大,而标准地法又存在着难于布设的问题。本研究提出的面积单元线性测量法,可方便快捷地进行局部估测,节省人力和时间,同时又达到了较高的精度,特别适用于不规则林分的调查。     

Chemical correlates of α-tocopherol (Vitamin E) alteredMalacosoma disstria herbivory inFraxinus pennsylvanica var.Subintegerrinia,green ash

The antioxidant -tocopherol (vitamin E), applied in a basal trunk band to the green ash tree,Fraxinus pennsylvanica var.subintergerrinia, elicited an alteration of foliar feeding byMalacosoma disstria larvae (Lepidoptera: Lasiocampidae). The bioassayed effects were dependent on the dosage of elicitor, the time after elicitation, and the position in the tree. Leaves for chemical analysis were collected from trees receiving two dosages and at two intervals after elicitation. Compounds in the ethyl acetate extractables from the ash tree leaves were separated by TLC and HPLC. TLC separations showed differences in the nonhydrolyzed extractables attributable to elicitor dosage and time after elicitation. TLC-resolved differences were also evident among acid-hydrolyzed samples. HPLC-resolved profiles revealed eight peaks in the nonhydrolyzed extractables that were quantitatively negatively correlated with larval feeding preference between elicited versus control foliage on at least one of the two sampling dates. Results from this study and other investigations reported in the literature indicate that the antioxidant-tocopherol (vitamin E) can function as an environmental-stress elicitor of alterable defensive chemistry in green ash and other plants.