关键词： Machine learning methods (ML methods)   Molecular descriptors   Recursive feature elimination (RFE)   Support vector machine (SVM)   Acetylcholinesterase (AChE)   Alzheimer's disease (AD)  

摘要： Acetylcholinesterase (AChE) has become an important drug target and its inhibitors have proved useful in the symptomatic treatment of Alzheimer's disease. This work explores several machine learning methods (support vector machine (SVM), k-nearest neighbor (k-NN), and C4.5 decision tree (C4.5 DT)) for predicting AChE inhibitors (AChEIs). A feature selection method is used for improving prediction accuracy and selecting molecular descriptors responsible for distinguishing AChEIs and non-AChEIs. The prediction accuracies are 76.3% similar to 88.0% for AChEIs and 74.3% similar to 79.6% for non-AChEIs based on the three kinds of machine learning methods. This work suggests that machine learning methods such as SVM are facilitating for predicting AChEIs potential of unknown sets of compounds and for exhibiting the molecular descriptors associated with AChEIs. (C) 2009 Elsevier Masson SAS. All rights reserved.

关键词： 支持向量机激   素敏感脂肪酶   机器学习方法   分子描述符   递归变量消除法  

摘要： 脂肪组织中,激素敏感脂肪酶(HSL)被认为是调节脂肪酸代谢的关键限速酶.HSL在糖尿病的发病过程中起重要作用,抑制HSL活性有助于糖尿病的治疗,因此探索新颖的HSL抑制剂成为当前研究的热门.在激素敏感脂肪酶的作用机制和三维结构缺乏的情况下,需要发展预测HSL抑制剂的方法.本文采用几种机器学习方法(支持向量机(SVM)、k-最近相邻法(k-NN)和C4.5决策树(C4.5DT))对已知的HSL抑制剂与非抑制剂建立分类预测模型.252个结构多样性化合物(123个HSL抑制剂与129个HSL非抑制剂)被用于测试分类预测系统,并用递归变量消除法选择与HSL抑制剂相关的性质描述符以提高预测精度.本研究对独立验证集的总预测精度为75.0%-80.0%,HSL抑制剂的预测精度为85.7%-90.5%,非HSL抑制剂的预测精度为63.2%-68.4%.支持向量机方法给出最好的总预测精度(80.0%).本研究表明支持向量机等机器学习方法可以有效预测未知数据集中潜在的HSL抑制剂,并有助于发现与其相关的分子描述符.

关键词： wearable sensors   accelerometers   motion analysis   human physical activity   machine learning   statistical pattern recognition   Hidden Markov Models  

摘要： The use of on-body wearable sensors is widespread in several academic and industrial domains. Of great interest are their applications in ambulatory monitoring and pervasive computing systems;here, some quantitative analysis of human motion and its automatic classification are the main computational tasks to be pursued. In this paper, we discuss how human physical activity can be classified using on-body accelerometers, with a major emphasis devoted to the computational algorithms employed for this purpose. In particular, we motivate our current interest for classifiers based on Hidden Markov Models (HMMs). An example is illustrated and discussed by analysing a dataset of accelerometer time series.

关键词： Metabolic Network   Enzyme Commission Number   Taxonomic Range   Pathway Prediction   Gold Standard Dataset  

摘要： Background: A key challenge in systems biology is the reconstruction of an organism's metabolic network from its genome sequence. One strategy for addressing this problem is to predict which metabolic pathways, from a reference database of known pathways, are present in the organism, based on the annotated genome of the organism. Results: To quantitatively validate methods for pathway prediction, we developed a large "gold standard" dataset of 5,610 pathway instances known to be present or absent in curated metabolic pathway databases for six organisms. We defined a collection of 123 pathway features, whose information content we evaluated with respect to the gold standard. Feature data were used as input to an extensive collection of machine learning (ML) methods, including naive Bayes, decision trees, and logistic regression, together with feature selection and ensemble methods. We compared the ML methods to the previous PathoLogic algorithm for pathway prediction using the gold standard dataset. We found that ML-based prediction methods can match the performance of the PathoLogic algorithm. PathoLogic achieved an accuracy of 91% and an F-measure of 0.786. The ML-based prediction methods achieved accuracy as high as 91.2% and F-measure as high as 0.787. The ML-based methods output a probability for each predicted pathway, whereas PathoLogic does not, which provides more information to the user and facilitates filtering of predicted pathways. Conclusions: ML methods for pathway prediction perform as well as existing methods, and have qualitative advantages in terms of extensibility, tunability, and explainability. More advanced prediction methods and/or more sophisticated input features may improve the performance of ML methods. However, pathway prediction performance appears to be limited largely by the ability to correctly match enzymes to the reactions they catalyze based on genome annotations.

关键词： Classification   Data complexity   Fuzzy rule based systems   Genetic fuzzy systems  

摘要： The analysis of data complexity is a proper framework to characterize the tackled classification problem and to identify domains of competence of classifiers. As a practical outcome of this framework, the proposed data complexity measures may facilitate the choice of a classifier for a given problem. The aim of this paper is to study the behaviour of a fuzzy rule based classification system and its relationship to data complexity. We use as a case of study the fuzzy hybrid genetic based machine learning method presented in [H. Ishibuchi, T. Yamamoto, T. Nakashima, Hybridization of fuzzy GBML approaches for pattern classification problems, IEEE Transactions on Systems, Man, and Cybemetics-Part B: Cybernetics 35 (2) (2005) 359-365]. We examine several metrics of data complexity over a wide range of data sets built from real data and try to extract behaviour patterns from the results. We obtain rules which describe both good or bad behaviours of the fuzzy rule based classification system. These rules use values of data complexity metrics in their antecedents, so we try to predict the behaviour of the method from the data set complexity metrics prior to its application. Therefore, we can establish the domains of competence of this fuzzy rule based classification system. (C) 2009 Elsevier B.V. All rights reserved.

关键词： 空间数据挖掘   L—M算法   可视化   分类预测  

摘要： 地理信息分类的传统线性算法具有正向直接判定的快速优势,但局限于对已知数据进行线性的判别划分,而非线性未知信息的分类预测同样是GIS技术的重要内容。人工神经网络算法为一些非线性知识的发现提供了可能。本文在通用的GIS格式数据基础上,采用L-M算法进行分类,通过分类结果来预测未知信息。开发出可视化的GIS数据神经网络分类预测软件模块。并以美国各镇人口为样例数据进行测试,分类预测结果显示该算法具有可行性及系统具有实用性。

关键词： binding affinity prediction   spherical self-organizing maps   support vector machines  

摘要： Machine learning methods have always been promising in the science and engineering fields, and the use of these methods in chemistry and drug design has advanced especially since the 1990s. In this study, molecular electrostatic potential (MEP) surfaces of phencyclidine-like (PCP-like) compounds are modeled and visualized in order to extract features that are useful in predicting binding affinities. In modeling, the Cartesian coordinates of MEP surface points are mapped onto a spherical self-organizing map (SSOM). The resulting maps are visualized using electrostatic potential (ESP) values. These values also provide features for a prediction system. Support vector machines and partial least-squares method are used for predicting binding affinities of compounds. Copyright (C) 2009 John Wiley & Sons, Ltd.

关键词： Sensor selection   Condition monitoring   Feature-group selection   Hierarchical Bayes   Empirical Bayes   Diesel engine fault detection  

摘要： In applications like feature-level sensor fusion, the problem of selecting an optimal number of sensors can lead to reduced maintenance costs and the creation of compact online databases for future use. This problem of sensor selection can be reduced to the problem of selecting an optimal set of groups of features during model selection. This is a more complex problem than the problem of feature selection, which has been recognized as a key aspect of statistical model identification. This work proposes a new algorithm based on the use of a Bayesian framework for the purpose of selecting groups of features during regression and classification. The hierarchical Bayesian formulation introduces grouping for the parameters of a generalized linear model and the model hyper-parameters are estimated using an empirical Bayes procedure. A novel aspect of the algorithm is its ability to simultaneously perform feature selection within groups to reduce over-fitting of the data. Further, the parameters obtained from this algorithm can be used to obtain a rank order among the selected sensors. The performance of the algorithm is first tested on a synthetic regression example. Finally, it is applied to the problem of fault detection in diesel engines (30,000 data records from 43 sensors, 8 classes) and used to compare the misclassification rates with a varying number of sensors. (C) 2009 Elsevier Ltd. All rights reserved.

关键词： 高光谱影像   机器学习   非负矩阵分解   广义判别分析   支持向量机   简约集   相关向量机   模糊分类   AdaBoost  

摘要： 高光谱影像分类是高光谱遥感应用的关键技术之一,对于资源调查、环境监测、精细农业、军事测绘与战场环境探测等应用具有重要意义。为了满足高光谱影像分类对精度、速度和可靠性的要求,针对高光谱影像具有的高维相关特征、非线性可分的特点,本文围绕影像分类及特征提取技术,采用机器学习方法展开了深入研究。论文的主要研究内容和创新点如下: 1.针对基于乘法迭代规则的高光谱影像非负矩阵分解(NMF)收敛速度慢的问题,给出了一种基于改进投影梯度的快速分解算法。采用分块坐标下降法将NMF的整体优化转化为两个子优化问题迭代求解,而每个子优化问题都通过改进的投影梯度法计算。实验分析表明,该方法能够明显提高高光谱影像NMF的收敛速度。 2.针对高光谱影像广义判别分析(GDA)特征提取中核函数参数选择问题,给出了一种高斯径向基核函数参数选择方法。将训练样本数值范围规范化,按对数形式将参数空间离散化,通过交叉验证获取合适参数。实验分析显示,使用该方法选择的参数,GDA特征提取能够显著提高高光谱影像分类的精度。 3.提出了一种基于简约集(RS)支持向量机(SVM)的高光谱影像分类方法。采用序列最小优化算法、交叉验证网格搜索参数选择方法构造高精度的多类SVM分类器,通过微分进化算法求解原像问题来获取简约集向量。经验证,SVM具有良好的泛化能力,RS-SVM能够保持分类精度,有效提高分类速度。 4.提出了一种基于相关向量机(RVM)的高光谱影像模糊分类方法。采用序列稀疏贝叶斯学习算法提高RVM训练速度,针对一对一法构造的多类RVM分类器,将两两配对的概率输出转化为相对于地物类别的隶属度。与SVM的比较表明,RVM参数选择简单、分类速度快;利用模糊隶属度能够标识混合像元,有效提高影像分类的可靠性。 5.提出了一种基于自适应提升(AdaBoost)高光谱影像集成分类方法。选择决策树树桩作为弱分类器,通过Gentle AdaBoost算法提升为分段线性的强分类器,采用一对余法解决多类分类问题。结果表明,AdaBoost分类方法的训练和分类速度快,分类精度优于一般分类方法。 6.通过理论和实验分析了SVM、RVM和AdaBoost在分类精度、训练和分类速度方面的表现,指出了各自在不同高光谱影像分类应用中的潜力。SVM适用于实时性要求不高的地物精细分类,RVM适用于获取具有统计意义的地物分类,而AdaBoost适用于分类精度要求较高的快速分类。

关键词： 攻击   防范   机器学习   语义   数据流分析   HMM   集成学习  

摘要： 近两年美国因病毒、间谍软件等网络攻击损失近85亿美元,而中国大陆更是有数以亿计的大量主机和网络被恶意攻击、破坏和篡改。一方面,种类繁多功能各异的诸如病毒,蠕虫,rootkit,间谍软件等恶意代码层出不穷,黑客攻击方式、手段与过程不断复杂深化;另一方面,信息系统漏洞不断增长,漏洞越来越多。强大的经济利益的驱动使得恶意代码检测与防范问题仍是信息安全届亟需解决的首要问题。攻击者不断升级并复杂化新的攻击手段,防守者根据攻击提出防护措施,例如修补漏洞,注入防护疫苗等;进一步地,攻击者提出新的反检测和规避技术,防护者也要不断更新防护技术。攻防双方不断博弈,两者在动态平衡中,不断将局部的马鞍点向前推进。 统计机器学习源于统计,长于关系推理和知识的自动学习,已在文本分析,视频分析,图像理解,语音信号识别取得极好的效果。我们把恶意代码检测与攻击比作一场猫捉老鼠的游戏,统计机器学习能不能有效的扮演“猫”的角色,能不能在已有的恶意代码检测与分析的基础上在如虎添翼?该问题的难点在于安全信息系统的一些特征需求与机器学习应用需求不是完全一致。例如信息安全中,对于误报率和漏报率的容忍度达到了苛刻的程度;对机器学习的结果缺少解释,模型的结果与实际的安全保障之间存在语义上的差距,很多结果在实际中不可行或者严重偏离信息系统程序和系统配置的现实;机器学习算法必须考虑攻击和攻击者各种各样的逃避检测策略。几乎信息安全的所有问题都是攻击者和防守者之间的博弈过程,必须站在双方的角度上着想,才有助于问题的解决。 针对代码分析的具体领域,在结合代码分析领域内知识的基础上,我们提出以下问题作为本文的研究对象。a)机器学习能不能在恶意代码或者代码分析中使用?b)在恶意代码检测（扩展到代码分析甚至系统安全中）,能起多大作用,如何使用并使其发挥最大功效?本文将此抽象问题具体化为几个子问题（Q1-Q4）进行细化,并通过具体的案例分析来回答。Q1：如何提取多态蠕虫签名?Q2：如何进行多态shellcode归属性分析?Q3：如何检测迷惑恶意代码?Q4：多线程程序中,如何消除时序相关的不确定性bug?本文关注的恶意代码包含两类,第一类是基于网络包的恶意代码,例如多态蠕虫,通过网络传播的shellcode;第二类是基于文件的恶意代码,例如被攻陷的可执行文件或者动态链接库文件;另外本文还分析了一个多线程程序安全中的案例。 针对上述问题,我们进行了下列研究：结合语义和统计特征,对多态蠕虫提取签名;结合语义和统计特征,对多态shellcode进行归属性分析;结合语义特征和统计特征,检测迷惑恶意代码;结合多线程运行的上下文,来推测时序对不确定性bug的影响。 我们的工作有以下创新点。a)提出了语义分析和统计分析相结合的代码分析新方法,用于检测或者分类恶意代码文件以及恶意代码包;与语义分析方法相比,融合了统计方法定量描述的特长;与统计方法相比,关注了更多的代码语义特性,使得分析更加接近代码语义本质。b)提出了基于数据流分析的状态转移图签名,用于多态蠕虫签名提取,通过数据流分析去除隐含在网络数据包中的噪声数据,较好刻画蠕虫的多态特性。c)提出了一种结合静态污点分析和混合Markov模型的shellcode归属性分析算法;通过静态污点分析保留语义相关字节,混合Markov模型获取数据包的统计结构特征;比单一的统计分析更加健壮,比仅仅的静态污点分析方法更易于定量描述和进行代码相似性比较。d)提出了一种结合控制流和系统调用特征的迷惑恶意代码检测算法,用于检测迷惑后的恶意代码;控制流和系统调用获取了代码的语义特征,而同时结合统计特征,相互补充,尽可能准确的进行迷惑恶意代码的类别检测。e)提出了使用HMM,刻画影响多线程运行的上下文,通过上下文（优先级,系统负载,运行时间等）捕获环境对程序运行的影响,并将这种影响进一步量化,为不确定性bug提供分析的依据。 通过以上研究,我们发现,结合或者体现某种程度语义的机器学习可以较为有效的应用于代码分析和检测中。结合语义的多态shellcode签名提取和归属性分析的相关技术已经出现在DayZeroSystems产品中,提升多维特征的迷惑恶意代码检测方法已开始应用到下一代智能安全检测产品中（例如Damballa,Inc）。我们相信这些技术将会在智能信息安全检测中得到更广泛应用。