关键词： 阿尔茨海默病   磁共振成像   人工智能   诊断  

摘要： 阿尔茨海默病（Alzheimer′s Disease,AD）是一种以认知和智力损害、行为能力下降为特点的神经退行性疾病,目前其确切病因不明,并缺少有效的治疗方案（[1]）。据估计,到2050年,AD的全球发病率将达1/85（[2]）。轻度认知功能障碍（mild cognitive impairment,MCI）是AD与健康老化的中间阶段,但并不是所有的MCI患者都会转化成AD.

关键词： PCI术预后   MACE   机器学习   非均衡分类问题  

摘要： 目的:通过对比分析六种典型机器学习方法构建的PCI术后12个月MACE(主要不良心血管事件)风险预测模型,获得最优MACE风险预测模型。方法:利用随机抽样和SMOTE方法进行样本均衡化处理,再分别利用CART回归树、C4.5条件树、Ada Boost、Bagging、Random Forest(RF)和Na?ve Bayes六种机器学习方法构建MACE风险预测模型。结果:SMOTE方法处理后的训练集可以提高预测模型的预测能力。利用RF算法构建的模型预测精度达到88.7%,预测能力明显优于其他模型。结论:利用RF算法构建的PCI术后发生MACE的风险预测模型,可有效帮助医护人员对患者提早进行干预指导,降低患者不良事件的发生。

关键词： Multi-criteria decision strategy   analytical hierarchy process   indoor positioning   fingerprinting   machine learning  

摘要： Indoor positioning system is an active research area. There are various performance metrics such as accuracy, computation time, precision, and f-score in machine learning based indoor positioning systems. The aim of this study is to present a multi-criteria decision strategy to determine suitable machine learning methods for a specific indoor positioning system. This helps to evaluate the performance of machine learning algorithms considering multiple criteria. During the experiments, UJllndoorLoc, KIOS and RFKON datasets are used from the positioning literature. The algorithms such as k-nearest neighbor, support vector machine, decision tree, naive bayes and bayesian networks are compared using these datasets. In addition to these, ensemble learning algorithms, namely adaboost and bagging, are utilized to improve the performance of these classifiers. As a conclusion, the test results for any specific dataset are reevaluated using the performance metrics such as accuracy, f-score and computation time, and a multi-criteria decision strategy is proposed to find the most convenient algorithm. The analytical hierarchy process is used for multi-criteria decision. To the best of our knowledge, this is the first work to select the proper machine learning algorithm for an indoor positioning system using multi-criteria decision strategy.

关键词： Cancer prediction   Cancer recurrence   Cancer relapse   Data mining   Machine learning   Electronic health records  

摘要： Advancements are constantly being made in oncology, improving prevention and treatment of cancers. To help reduce the impact and deadliness of cancers, they must be detected early. Additionally, there is a risk of cancers recurring after potentially curative treatments are performed. Predictive models can be built using historical patient data to model the characteristics of patients that developed cancer or relapsed. These models can then be deployed into clinical settings to determine if new patients are at high risk for cancer development or recurrence. For large-scale predictive models to be built, structured data must be captured for a wide range of diverse patients. This paper explores current methods for building cancer risk models using structured clinical patient data. Trends in statistical and machine learning techniques are explored, and gaps are identified for future research. The field of cancer risk prediction is a high-impact one, and research must continue for these models to be embraced for clinical decision support of both practitioners and patients.

关键词： Perovskite materials   Curie temperature   Machine learning   Support vector machine   Relevance vector machine   Random forest  

摘要： Curie temperature (Tc), the second order phase transition temperature, is also one of the important physical properties of perovskite materials. It is a meaningful work to quickly and efficiently predict Tc of new perovskite materials before doing a considerable amount of experimental work. In the work, SVM (support vector machine), RVM (relevance vector machine) and RF (random forest) were employed to establish the prediction models of Tc with the physicochemical parameters, respectively. The results reveal that the three models all have high precision and reliability. According to K-fold cross validation, the SVR model had better prediction performance than the RVM and RF models. Meanwhile, the potential perovskite material with higher Tc was found by using the SVR model integrated with the search strategy of genetic algorithm from the virtual samples. The methods outlined here can provide valuable hints into the exploration of materials with desired property and can accelerate the process of materials design.

关键词： machine learning methods   St. Venant equations   rating curve method   nonlinear Muskingum model   hydrograph predictions  

摘要： Machine learning (soft) methods have a wide range of applications in many disciplines, including hydrology. The first application of these methods in hydrology started in the 1990s and have since been extensively employed. Flood hydrograph prediction is important in hydrology and is generally done using linear or nonlinear Muskingum (NLM) methods or the numerical solutions of St. Venant (SV) flow equations or their simplified forms. However, soft computing methods are also utilized. This study discusses the application of the artificial neural network (ANN), the genetic algorithm (GA), the ant colony optimization (ACO), and the particle swarm optimization (PSO) methods for flood hydrograph predictions. Flow field data recorded on an equipped reach of Tiber River, central Italy, are used for training the ANN and to find the optimal values of the parameters of the rating curve method (RCM) by the GA, ACO, and PSO methods. Real hydrographs are satisfactorily predicted by the methods with an error in peak discharge and time to peak not exceeding, on average, 4% and 1%, respectively. In addition, the parameters of the Nonlinear Muskingum Model (NMM) are optimized by the same methods for flood routing in an artificial channel. Flood hydrographs generated by the NMM are compared against those obtained by the numerical solutions of the St. Venant equations. Results reveal that the machine learning models (ANN, GA, ACO, and PSO) are powerful tools and can be gainfully employed for flood hydrograph prediction. They use less and easily measurable data and have no significant parameter estimation problem.

关键词： Ransomware   Volatile memory   Forensics   Memory dumps   Virtual machine   Private cloud   Machine Learning   Detection   Malware  

摘要： Cloud computing is one of today's most popular and important IT trends. Currently, most organizations use cloud computing services (public or private) as part of their computer infrastructure. Virtualization technology is at the core of cloud computing, and virtual resources, such as virtual servers, are commonly used to provide services to the entire organization. Due to their importance and prevalence, virtual servers in an organizational cloud are constantly targeted by cyber-attackers who try to inject malicious code or malware into the server (e.g., ransomware). Many times, server administrators are not aware that the server has been compromised, despite the presence of detection solutions on the server (e.g., antivirus engine). In other cases, the breach is detected after a long period of time when significant damage has already occurred. Thus, detecting that a virtual server has been compromised is extremely important for organizational security. Existing security solutions that are installed on the server (e.g., antivirus) are considered untrusted, since malware (particularly sophisticated ones) can evade them. Moreover, these tools are largely incapable of detecting new unknown malware. Machine learning (ML) methods have been shown to be effective at detecting malware in various domains. In this paper, we present a novel methodology for trusted detection of ransomware in virtual servers on an organization's private cloud. We conducted trusted analysis of volatile memory dumps taken from a virtual machine (memory forensics), using the Volatility framework, and created general descriptive meta-features. We leveraged these meta-features, using machine learning algorithms, for the detection of unknown ransomware in virtual servers. We evaluated our methodology extensively in five comprehensive experiments of increasing difficulty, on two different popular servers (IIS server and an email server). We used a collection of real-world, professional, and notorious r

关键词： Radon   Machine learning   Neural network   Linear regression   Environment   Sensitivity analysis  

摘要： Machine learning is a class of statistical techniques which has proven to be a powerful tool for modelling the behaviour of complex systems, in which response quantities depend on assumed controls or predictors in a complicated way. In this paper, as our first purpose, we propose the application of machine learning to reconstruct incomplete or irregularly sampled data of time series indoor radon (Rn-222). The physical assumption underlying the modelling is that Rn concentration in the air is controlled by environmental variables such as air temperature and pressure. The algorithms "learn" from complete sections of multivariate series, derive a dependence model and apply it to sections where the controls are available, but not the response (Rn), and in this way complete the Rn series. Three machine learning techniques are applied in this study, namely random forest, its extension called the gradient boosting machine and deep learning. For a comparison, we apply the classical multiple regression in a generalized linear model version. Performance of the models is evaluated through different metrics. The performance of the gradient boosting machine is found to be superior to that of the other techniques. By applying learning machines, we show, as our second purpose, that missing data or periods of Rn series data can be reconstructed and resamplecl on a regular grid reasonably, if data of appropriate physical controls are available. The techniques also identify to which degree the assumed controls contribute to imputing missing Rn values. Our third purpose, though no less important from the viewpoint of physics, is identifying to which degree physical, in this case environmental variables, are relevant- as Rn predictors, or in other words, which predictors explain most of the temporal variability of Rn. We show that variables which contribute most to the Rn series reconstruction, are temperature, relative humidity and day of the year. The first two are physical predictor

关键词： Animal cells   Artificial intelligence   Artificial neural networks   Asia   Biology and life sciences   Cell biology   Cellular neuroscience   Cellular types   China   Computational biology   Computational neuroscience   Computer and information sciences   Computer networks   Database and informatics methods   Epidemiology   Forecasting   Geographical locations   HIV epidemiology   Infectious diseases   Information retrieval   Internet   Mathematical and statistical techniques   Mathematics   Medicine and health sciences   Neurons   Neuroscience   People and places   Physical sciences   Research and analysis methods   Research Article   Statistical methods   Statistics (mathematics)   Viral diseases  

摘要： Background AIDS is a worrying public health issue in China and lacks timely and effective surveillance. With the diffusion and adoption of the Internet, the 'big data' aggregated from Internet search engines, which contain users' information on the concern or reality of their health status, provide a new opportunity for AIDS surveillance. This paper uses search engine data to monitor and forecast AIDS in China. Methods A machine learning method, artificial neural networks (ANNs), is used to forecast AIDS incidences and deaths. Search trend data related to AIDS from the largest Chinese search engine, Baidu. com, are collected and selected as the input variables of ANNs, and officially reported actual AIDS incidences and deaths are used as the output variable. Three criteria, the mean absolute percentage error, the root mean squared percentage error, and the index of agreement, are used to test the forecasting performance of the ANN method. Results Based on the monthly time series data from January 2011 to June 2017, this article finds that, under the three criteria, the ANN method can lead to satisfactory forecasting of AIDS incidences and deaths, regardless of the change in the number of search queries. Conclusions Despite the inability to self-detect HIV/AIDS through online searching, Internet-based data should be adopted as a timely, cost-effective complement to a traditional AIDS surveillance system.

关键词： 机器学习   敏感性评估   K近邻   朴素贝叶斯   随机森林  

摘要： 为了比较不同机器学习方法在灾害敏感性评估中的适用性,将K近邻、朴素贝叶斯和随机森林3种典型机器学习方法运用到灾害敏感性评估中,比较3种方法在预测准确率、混淆矩阵和敏感性图上的异同;并基于深圳市电动自行车出险数据以及孕灾环境数据,运用3种方法评估深圳市电动自行车出险的敏感性,进而绘制敏感性图,探讨3种方法对于不同性质数据的适用条件,为开展灾害敏感性评估工作时选取更为适宜的方法,提供参考建议。