关键词： Neural networks  

摘要： In the past decade, machine learning methods for empirical rainfall-runoff modeling have seen extensive development and been proposed as a useful complement to physical hydrologic models, particularly in basins where data to support process-based models are limited. However, the majority of research has focused on a small number of methods, such as artificial neural networks, despite the development of multiple other approaches for non-parametric regression in recent years. Furthermore, this work has often evaluated model performance based on predictive accuracy alone, while not considering broader objectives, such as model interpretability and uncertainty, that are important if such methods are to be used for planning and management decisions. In this paper, we use multiple regression and machine learning approaches (including generalized additive models, multivariate adaptive regression splines, artificial neural networks, random forests, and M5 cubist models) to simulate monthly streamflow in five highly seasonal rivers in the highlands of Ethiopia and compare their performance in terms of predictive accuracy, error structure and bias, model interpretability, and uncertainty when faced with extreme climate conditions. While the relative predictive performance of models differed across basins, data-driven approaches were able to achieve reduced errors when compared to physical models developed for the region. Methods such as random forests and generalized additive models may have advantages in terms of visualization and interpretation of model structure, which can be useful in providing insights into physical watershed function. However, the uncertainty associated with model predictions under extreme climate conditions should be carefully evaluated, since certain models (especially generalized additive models and multivariate adaptive regression splines) become highly variable when faced with high temperatures.

关键词： 冠心病   RBF-DDA神经网络   支持向量机   随机森林   慢性病筛查  

摘要： 目的:针对慢性病调查数据,建立RBF-DDA神经网络、支持向量机、随机森林三种冠心病识别模型,探索不同机器学习方法在冠心病筛查识别中的应用,验证多种机器学习方法在冠心病筛查中应用的可能性,为冠心病筛查方法的改进提供支持。方法:将2012年开展的吉林省成人慢性病及其危险因素调查的数据划分为训练数据集和测试数据集。首先利用训练数据集建立RBF-DDA神经网络模型、支持向量机模型和随机森林模型的冠心病筛查识别模型,利用测试数据集的输入变量输入已经建立好的模型,预测识别输出变量,并与测试数据集中的实际输出变量进行对比,产生模型的识别效果。然后通过手工、十折交叉验证等方式对不同模型中涉及的参数进行优化。最后,利用最优参数建立三种模型,对模型的筛查识别效果进行检验,并通过精度、灵敏度、特异度等指标对不同模型的筛查识别效果进行比较。结果:***-DDA神经网络取得的最优参数为:激活阈值(0.5)和抑制阈值(0.5)。将RBF-DDA神经网络模型在测试集中的是否患冠心病预测值与实际值进行比较,模型的精度55.83%、灵敏度58.88%、特异度55.46%、G-Mean57.14%。RBF-DDA神经网络模型的结果优于传统RBF神经网络。2.支持向量机有四种不同的核函数,不同核函数建立四种支持向量机模型:linear-SVM、radial-SVM、sigmoid-SVM、polynomial-SVM。linear-SVM模型的最优参数为,cost=0.5;radial-SVM模型的最优参数为,cost=2、gamma=0.01;sigmoidSVM模型的最优参数为,cost=4、gamma=0.001、coef0=0.25;polynomial-SVM模型的最优参数为,cost=8、gamma=0.001、coef0=0.25、degree=3。经过参数优化后,得到四种模型的最优识别效果。linear-SVM模型的精度68.03%、灵敏度76.45%、特异度67.01%、G-Mean71.57%。radial-SVM模型的精度65.32%、灵敏度77.57%、特异度63.84%、G-Mean70.37%。sigmoid-SVM模型的精度67.93%、灵敏度77.94%、特异度66.71%、G-Mean72.11%。polynomial-SVM模型的精度67.58%、灵敏度79.07%、特异度66.19%、G-Mean72.34%。3.随机森林模型采用手工和十折交叉验证两种参数优化方式。手工参数优化取得的最优参数为:mtry=6、ntree=300。十折交叉验证参数优化取得的最优参数为:mtry=6、ntree=290。手工优化的随机森林模型的精度66.86%、灵敏度81.50%、特异度65.08%、G-Mean72.83%。十折交叉验证优化的随机森林模型的精度66.49%、灵敏度80.56%、特异度64.79%、G-Mean72.24%。4.三种模型识别效果的比较,随机森林的综合效果最好,模型的精度66.86%、灵敏度81.50%、特异度65.08%、G-Mean72.83%。在精度和特异度上,随机森林与支持向量机相差不大,且均比RBF-DDA神经网络模型要好。在灵敏度上,随机森林最好,比支持向量机高2.43%,比RBF-DDA神经网络高22.62%。在综合指标G-Mean上,随机森林比支持向量机高0.49%,比RBF-DDA神经网络高15.69%。结论:***-DDA神经网络、支持向量机和随机森林三种模型在基于慢性病调查资料的冠心病识别运用中具有可行性。2.针对模型参数的优化,对RBF-DDA神经网络、支持向量机和随机森林三个模型的冠心病识别效果均有提升。3.三个模型的识别效果:随机森林模型最好、支持向量机其次、RBFDDA神经网络最差。

关键词： 统计建模方法   随机森林   mboost   神经网络   交叉验证   标准均方误差  

摘要： 本文旨在用近年较经典流行的三种机器学习算法:随机森林,神经网络,mboost与统计建模方法对多元时间序列数据(气温数据)、纵向数据(帕金森病数据)、多重共线性数据(糖尿病数据)做预测对比.针对世界16个城市的最低温度和最高温度为变量的多元时间序列数据,运用VARX模型预测和四种主要的针对横截面数据的机器学习算法做预测比较,并作出对比分析,结果表明无论是在长期预测还是短期预测,VARX模型的预测效果大都不如这些机器学习算法中的一些模型.对于对各个变量中选择出来的不同因变量,预测结果最优的模型也不同.被预测的时间区间越短(相应的训练集越大),则VARX模型预测的效果越好,这说明专门为时间序列设计的方法在该数据上的预测不如为横截面数据设计的机器学习算法,因而在做多元时间序列预测的时候,必须根据数据和变量的各种实际情况来选择最优的预测模型.对一个帕金森病的纵向数据和一个糖尿病的横截面数据做了机器学习方法及统计建模方法对预测的比较.关于第一个数据,对于训练集不同的样本量,分别用线性随机效应混合模型和随机森林、mboost、神经网络做了预测,并对比标准化均方误差.发现无论是长期预测还是短期预测,线性随机效应混合模型在该数据上的预测效果都显著不如神经网络.对于第二个数据:具有多重共线性,分别用岭回归,lasso回归,适应性lasso回归,偏最小二乘回归(PLS),逐步回归,线性回归及机器学习算法做十折交叉验证预测对比.结果显示,神经网络在处理此多重共线性数据时远远好于其他的传统统计建模方法,而不那么传统的PLS方法也全面优于其它几种传统方法,但远不如神经网络方法.本文为比较研究的案例库贡献新的、有重要参考价值的比较案例,同时对实际工作者提供有益的参考.本文的所有计算基于R软件.

关键词： Carbonation   Concrete   Machine learning   Neural network   Decision tree   Bagged decision tree   Boosted decision tree   Model  

摘要： Reliable carbonation depth prediction of concrete structures is crucial for optimizing their design and maintenance. The challenge of conventional carbonation prediction models is capturing the complex relationship between governing parameters. To improve the accuracy and methodology of the prediction a machine learning based carbonation prediction model which integrates four learning methods is introduced. The model developed considers parameters influencing the carbonation process and enables the user to choose the best alternative of the machine based methods. The applicability of the method is demonstrated by an example where the carbonation depths are estimated using the developed model and verified with unseen data. The evaluation proofs that the model predicts the carbonation depth with a high accuracy. (C) 2015 Elsevier Ltd. All rights reserved.

关键词： O-GlcNAcylation   O-linked glycosylation   O-GlcNAc transferase (OGT)   substrate motif   profile hidden Markov model   support vector machine  

摘要： Protein O-GlcNAcylation, involving the b-attachment of single N-acetylglucosamine (GlcNAc) to the hydroxyl group of serine or threonine residues, is an O-linked glycosylation catalyzed by O-GlcNAc transferase (OGT). Molecular level investigation of the basis for OGT's substrate specificity should aid understanding how O-GlcNAc contributes to diverse cellular processes. Due to an increasing number of O-GlcNAcylated peptides with site-specific information identified by mass spectrometry (MS)-based proteomics, we were motivated to characterize substrate site motifs of O-GlcNAc transferases. In this investigation, a non-redundant dataset of 410 experimentally verified O-GlcNAcylation sites were manually extracted from dbOGAP, OGlycBase and UniProtKB. After detection of conserved motifs by using maximal dependence decomposition, profile hidden Markov model (profile HMM) was adopted to learn a first-layered model for each identified OGT substrate motif. Support Vector Machine (SVM) was then used to generate a second-layered model learned from the output values of profile HMMs in first layer. The two-layered predictive model was evaluated using a five-fold cross validation which yielded a sensitivity of 85.4%, a specificity of 84.1%, and an accuracy of 84.7%. Additionally, an independent testing set from PhosphoSitePlus, which was really non-homologous to the training data of predictive model, was used to demonstrate that the proposed method could provide a promising accuracy (84.05%) and outperform other O-GlcNAcylation site prediction tools. A case study indicated that the proposed method could be a feasible means of conducting preliminary analyses of protein O-GlcNAcylation and has been implemented as a web-based system, OGTSite, which is now freely available at http://***/OGTSite/.

关键词： 磁盘   故障预测   机器学习  

摘要： 磁盘是保存数据的重要载体,提高磁盘的可靠性和数据可用性具有重要意义。现代磁盘普遍支持SMART协议,用来监控磁盘的内部工作状态。采用机器学习方法,分析磁盘的SMART信息,实现对磁盘故障的预测。所采用的机器学习方法包括反向神经网络、决策树、支持向量机以及简单贝叶斯,并采用实际磁盘SMART数据进行验证与分析。基于上述数据,对不同机器学习方法的有效性及其效果进行了对比。结果表明,决策树方法的预测率最好,支持向量机方法的误报率最低。

关键词： 机器学习方法   入侵检测技术   贝叶斯分类神经网络   遗传算法   支持向量机  

摘要： 随着信息技术的不断发展,人们生活发生了翻天覆地的变化,它给人们带来了很多便利,但与此同时,安全问题也日益突出。目前,传统的入侵检测系统已经不足以完成对越来越复杂的网络攻击的检测任务。入侵监测系统技术之中引入机器学习,可以有效地提高系统性能。文章主要介绍了几种机器学习方法在入侵检测中的应用。

关键词： Breast ultrasound   Breast cancer   Adaptive boosting   Computer-aided diagnosis   Artificial intelligence  

摘要： The goal of this study was to devise a machine learning methodology as a viable low-cost alternative to a second reader to help augment physicians' interpretations of breast ultrasound images in differentiating benign and malignant masses. Two independent feature sets consisting of visual features based on a radiologist's interpretation of images and computer-extracted features when used as first and second readers and combined by adaptive boosting (AdaBoost) and a pruning classifier resulted in a very high level of diagnostic performance (area under the receiver operating characteristic curve = 0.98) at a cost of pruning a fraction (20%) of the cases for further evaluation by independent methods. AdaBoost also improved the diagnostic performance of the individual human observers and increased the agreement between their analyses. Pairing AdaBoost with selective pruning is a principled methodology for achieving high diagnostic performance without the added cost of an additional reader for differentiating solid breast masses by ultrasound. (C) 2015 World Federation for Ultrasound in Medicine & Biology.

关键词： remote sensing   soil moisture   biomass   retrieval algorithms   machine learning   artificial neural networks   SVM   regression   biophysical parameters  

摘要： The enormous increase of remote sensing data from airborne and space-borne platforms, as well as ground measurements has directed the attention of scientists towards new and efficient retrieval methodologies. Of particular importance is the consideration of the large extent and the high dimensionality (spectral, temporal and spatial) of remote sensing data. Moreover, the launch of the Sentinel satellite family will increase the availability of data, especially in the temporal domain, at no cost to the users. To analyze these data and to extract relevant features, such as essential climate variables (ECV), specific methodologies need to be exploited. Among these, greater attention is devoted to machine learning methods due to their flexibility and the capability to process large number of inputs and to handle non-linear problems. The main objective of this paper is to provide a review of research that is being carried out to retrieve two critically important terrestrial biophysical quantities (vegetation biomass and soil moisture) from remote sensing data using machine learning methods.

关键词： 转录因子   转录因子家族分类   转录因子预测   植物转录因子数据库  

摘要： 转录因子在植物生长发育和应对胁迫等过程中具有重要调控作用,基因组水平上系统预测植物转录因子是研究其功能和演化的基础。通过深入全面的文献调研,总结了一套完整的植物转录因子家族分类规则,开发了转录因子预测流程,构建了转录因子预测平台。基于该流程,从83种绿色植物中预测到129 288个转录因子,分属58个家族。对预测到的转录因子,从家族和个体两个层次进行了详尽注释,构建了植物转录因子数据库Plant TFDB(http://***)。Plant TFDB提供了覆盖绿色植物主要谱系的转录因子,其中大部分为具有重要经济价值的单子叶和双子叶植物,已成为植物转录因子功能和演化研究的重要信息资源和分析平台。