关键词： Random forest   Classification   Vegetation index  

摘要： Growth and production models can help to simulate the growth of tree dimensions to predict forest productivity at different levels. In this context, the following questions arise: (i) is it possible to recognize the growth pattern of eucalyptus species based on spectral features using machine learning (ML) for data modeling? (ii) what spectral features provides better accuracy? and (iii) what ML algorithms are most accurate for performing this modeling? To answer these questions, the present study evaluated the use of ML techniques using breast height and total plant height to classify the growth of five species of eucalyptus and Corymbria citriodora in an unsupervised learning, and the obtained classes for induce ML algorithms to recognize the species with relation to their growth using vegetation indices (VIs) and spectral bands (SBs). It were evaluated five eucalyptus species (E. camaldulensis, E. uroplylla, E. saligna, E. grandis e E. urograndis) and C. citriodora in experimental design of randomized blocks with four replicates, with 20 plants inside each experimental plot. The diameter at breast height and total plant height at stand level were obtained by measuring five trees in each experimental unit in seven measurements. During this same period, a flight was carried out using a remotely piloted aircraft for the acquisition of spectral variables (SBs and VIs). For recognition of eucalyptus species in relation to their growth two machine learning approaches were employed: supervised and unsupervised. The average accuracy obtained from 10-fold cross-validation, employing Random Forest algorithm and 24 features, was 0.76. This result shows that the proposed approach is appropriate to recognize different eucalyptus species based on their growth.

关键词： 公交   多源数据   客流分类   卷积神经网络   门控制循环单元   组合预测模型  

摘要： 为精准分析公交线路与站点不同客流的出行特征及时变差异性,结合深度学习理论,提出了一种基于卷积神经网络(CNN)与门控制循环单元(GRU)组合的公交客流分类预测模型;融合匹配公交一卡通刷卡、公交车GPS轨迹、线路和站点基础信息、气象等多源数据,实现公交客流数据重构;采用K-Medians算法将乘客分为通勤类和非通勤类;以乘客类型、历史客流量、时段、高/平峰、星期、降水量、重大活动等因素为输入向量,分别建立CNN与GRU单一模型,并利用均方误差、均方根误差、平均绝对误差为评价指标,开展预测;针对单一模型不适用多特征时间序列预测等问题,分别构建了由CNN和GRU组合的线路客流和断面客流预测模型;以北京市特15路公交为例,预测工作日与非工作日场景下的线路及断面的分类客流。分析结果表明:对于通勤类和非通勤类线路及断面客流,组合模型的均方误差相比单一模型平均降低了57.932、13.106和33.987,均方根误差平均降低了1.862、1.058和1.538,平均绝对误差平均降低了1.399、0.487和0.613,可见,多源数据驱动下的CNN-GRU组合模型具有良好的预测性能。

关键词： 条件高斯贝叶斯网络   分类判别   代谢组学  

摘要： 目的探讨条件高斯贝叶斯网络(conditional Gaussian Bayesian network, CGBN)在代谢组学数据的分类判别中的应用。方法通过模拟研究与实际代谢组学数据分析,比较CGBN与偏最小二乘判别分析(partial least squares discriminant analysis, PLSDA)在不同相关程度和不同稀疏水平的高维数据及线性相关与非线性等情形时的分类判别性能,评价指标采用ROC曲线下面积(area under curve, AUC)和平均计算时间。结果模拟研究结果表明,变量之间低相关且样本量不大于200时CGBN分类判别AUC高于PLSDA。在自变量与因变量非线性相关且小样本情况下CGBN分类判别AUC同样高于PLSDA。实例数据分析结果显示CGBN和PLSDA分类判别的AUC分别为0.997,0.975。CGBN的计算时间要远高于PLSDA。结论在不受计算负担限制的情形下,CGBN是代谢组学数据典型分析方法的一种可行的替代方法,值得进一步研究。

关键词： 烃源岩   TOC   机器学习   CART  

摘要： 总有机碳(TOC)是烃源岩评价的关键参数,但烃源岩地化测试受取芯、测试费用等因素的限制,难以对工区烃源岩做出整体评价。主要提出了一种基于机器学习技术预测TOC的方法,是利用分类与回归树(CART)算法对测井曲线与实测TOC数据进行拟合,同时引入转化因子的倒数作为输入变量,来消除干酪根类型对计算结果的影响,建立CART预测TOC模型。结果表明,该模型训练集平均相对误差为4.39%,测试集平均相对误差为5.033%,与岩样测试结果吻合度高,可以进行推广。

关键词： 三疣梭子蟹   梯度提升回归树(GBRT)   支持向量机(SVM)   资源量  

摘要： 为了解东海北部海域三疣梭子蟹(Portunus trituberculatus)资源时空分布规律,探索更适合三疣梭子蟹资源量预测的模型方法,根据2006-2007年共四个季度在东海北部海域的底拖网调查数据,运用梯度提升回归树(gradient boosting regression tree,GBRT)和支持向量机(support vector machine,SVM)这两种机器学习方法,分析了三疣梭子蟹时空分布与环境因子之间的关系,同时使用方差解释率(VE)、相对均方根误差(RMSE)以及决定系数R2等指标对不同模型的拟合效果、预测性能以及稳定性等进行了比较,选择其中最佳模型对东海北部海域三疣梭子蟹资源分布进行预测。结果显示,GBRT模型的拟合效果相对优于SVM模型,两种模型的拟合结果均显示底层海水盐度(SBS)为影响三疣梭子蟹资源分布最为显著的环境因子。GBRT模型的预测性能较高且模型较为稳定,其预测结果显示夏季的资源量高于其他三个季节,且各季节所研究海域的东南部均存在一个资源分布的低值区。研究结果预期可为三疣梭子蟹资源分布及资源量预测新方法的探索和分析提供技术指导。

关键词： Chlorophyll-a   Machine learning   Menor sea   Sentinel 2   Water quality  

摘要： The Menor sea is a coastal lagoon declared by the European Union as a sensitive area to eutrophication due to human activities. To control the deterioration of its water quality, it is necessary to monitor some parameters such as chlorophyll-a (chl-a), which indicates phytoplankton biomass in the water. In the study area, current efforts focus on in-situ measurements to estimate chl-a by means of a few permanent stations and seasonal oceanographic campaigns, however they are expensive and time consuming. In this work, we proposed a machine learning approach based on Sentinel-2 data to estimate chl-a content on the upper part of the water column. Random forest (rf), support vector machine (svmRadial), Artificial Neural Network (ANN) and Deep Neural Network (DNN) algorithms were utilized under three feature selection scenarios, and several spectral indices were used in combination with Sentinel 2 bands. Rf, svmRadial and DNN performed better when all the available predictors were included in the models (RMSE = 0.82, 0.82 and 1.76 mg/m3 respectively), whereas ANN achieved better results under scenario c (principal components). Our results demonstrate the possibility to estimate chl-a concentration in a cost-effective manner and thereby provide near-real time information to monitor the water quality of the Menor sea, what can be of great interest for local authorities, tourism and fishing industry.

关键词： 机器学习   CT冠状动脉成像   血流储备分数   斑块特征  

摘要： 目的利用机器学习(ML)方法探讨定量、定性的冠状动脉斑块特征以及血流动力学参数对缺血特异性狭窄血管的影响。材料与方法回顾性分析同时行冠状动脉CT血管成像(CCTA)、有创冠状动脉造影(ICA)及FFR测量的255例病人共328支血管的临床及影像资料。以FFR≤0.80作为提示病变特异性缺血的指标,依据FFR值将纳入血管分为非缺血组(FFR>0.80)和缺血组(FFR≤0.80)。测量所有纳入斑块的CCTA参数,包括斑块的定量、定性特征以及血流动力学参数。采用LogitBoost算法建立随机森林模型,通过信息增益排序方法自动选择特征。分类准确度、敏感度以及f1值(分类准确度与敏感度的调和平均值)用以评价随机森林模型对造成缺血特异性狭窄斑块的预测价值,并计算受试者操作特征(ROC)曲线下面积。采用十折分层交叉验证法计算模型的总体分类准确度。结果应用ML方法得出,血流动力学参数FFR_(CT)、ΔFFR_(CT)是预测缺血特异性狭窄最重要的2个特征,其次是斑块的定量、定性特征,包括脂质斑块体积、斑块弯曲、斑块不规则、非钙化斑块体积、狭窄程度、纤维斑块体积和管腔体积。在排序前10的特征中有9个是CCTA相关参数,只有1个临床参数。采用随机森林模型预测缺血特异性狭窄的分类准确度为0.940,敏感度为0.940,f1值为0.940;预测缺血特异性狭窄的ROC曲线下面积为0.992,模型的总体分类准确度为0.921±0.047。结论ML方法能够很好地预测引起心肌缺血的冠状动脉特异性狭窄病变的斑块特征。

关键词： 多元线性回归   BP神经网络   随机森林   中长期潜势预报  

摘要： 以成都市为例,以多项可能影响污染物时空分布的变量为潜在预报因子,筛选关键入模因子,利用2016—2018年数据为训练集,采用多元线性回归、BP神经网络和随机森林算法,建立成都市夏季(4—8月)臭氧及冬季(11—2月)PM_(2.5)污染潜势模型,并利用2019年数据对模型的中长期污染潜势浓度的预报性能进行评估。结果表明,建立的多元线性回归、BP神经网络和随机森林模型对成都市臭氧及PM_(2.5)的短期(1~3天)污染潜势都具有良好的预报效果,对7~15天的中长期潜势预报表现稳定。其中,多元线性回归模型和随机森林模型分别对臭氧和PM_(2.5)表现出相对最佳的预报性能。

关键词： Amaranthus   chemometrics   Vis-NIR spectroscopy   machine learning   species discrimination   weeds  

摘要： The feasibility of rapid and non-destructive classification of six different Amaranthus species was investigated using visible-near-infrared (Vis-NIR) spectra coupled with chemometric approaches. The focus of this research would be to use a handheld spectrometer in the field to classify six Amaranthus sp. in different geographical regions of South Korea. Spectra were obtained from the adaxial side of the leaves at 1.5 nm intervals in the Vis-NIR spectral range between 400 and 1075 nm. The obtained spectra were assessed with four different preprocessing methods in order to detect the optimum preprocessing method with high classification accuracy. Preprocessed spectra of six Amaranthus sp. were used as input for the machine learning-based chemometric analysis. All the classification results were validated using cross-validation to produce robust estimates of classification accuracies. The different combinations of preprocessing and modeling were shown to have a classification accuracy of between 71% and 99.7% after the cross-validation. The combination of Savitzky-Golay preprocessing and Support vector machine showed a maximum mean classification accuracy of 99.7% for the discrimination of Amaranthus sp. Considering the high number of spectra involved in this study, the growth stage of the plants, varying measurement locations, and the scanning position of leaves on the plant are all important. We conclude that Vis-NIR spectroscopy, in combination with appropriate preprocessing and machine learning methods, may be used in the field to effectively classify Amaranthus sp. for the effective management of the weedy species and/or for monitoring their food applications.

关键词： radiomics   magnetic resonance imaging   multiple myeloma   cytogenetics   classifier  

摘要： Background Radiomics has shown promising results in the diagnosis, efficacy, and prognostic assessments of multiple myeloma (MM). However, little evidence exists on the utility of radiomics in predicting a high-risk cytogenetic (HRC) status in MM. Purpose To develop and test a magnetic resonance imaging (MRI)-based radiomics model for predicting an HRC status in MM patients. Study Type Retrospective. Population Eighty-nine MM patients (HRC [n: 37] and non-HRC [n: 52]). Field Strength/Sequence A 3.0 T;fast spin-echo (FSE): T1-weighted image (T1WI) and fat-suppression T2WI (FS-T2WI). Assessment Overall, 1409 radiomics features were extracted from each volume of interest drawn by radiologists. Three sequential feature selection steps-variance threshold, SelectKBest, and least absolute shrinkage selection operator-were repeated 10 times with 5-fold cross-validation. Radiomics models were constructed with the top three frequency features of T1WI/T2WI/two-sequence MRI (T1WI and FS-T2WI). Radiomics models, clinical data (age and visually assessed MRI pattern), or radiomics combined with clinical data were used with six classifiers to distinguish between HRC and non-HRC statuses. Six classifiers used were support vector machine, random forest, logistic regression (LR), decision tree, k-nearest neighbor, and XGBoost. Model performance was evaluated with area under the curve (AUC) values. Statistical Tests Mann-Whitney U-test, Chi-squared test, Z test, and DeLong method. Results The LR classifier performed better than the other classifiers based on different data (AUC: 0.65-0.82;P < 0.05). The two-sequence MRI models performed better than the other data models using different classifiers (AUC: 0.68-0.82;P < 0.05). Thus, the LR two-sequence model yielded the best performance (AUC: 0.82 +/- 0.02;sensitivity: 84.1%;specificity: 68.1%;accuracy: 74.7%;P < 0.05). Conclusion The LR-based machine learning method appears superior to other classifier methods for assessing HRC in MM. Ra