基于GA-BP神经网络模型的登革热时空扩散模拟

目的 登革热是一个全球性公共卫生问题,从地理学时空数据分析的视角,探究登革热的时空特质、构建登革热时空过程模型,是有效预防、控制登革热的新方法、研究新热点。方法 基于时空数据挖掘、时空过程建模,综合环境、气象、地理、人口4大因素,分析登革热的空间相关性及登革热病例的空间自相关,挖掘登革热影响因子;针对BP(back propagation)神经网络模型易陷入局部最优的缺陷,引入遗传算法(GA)改进BP神经网络模型,用于登革热时空模拟。结果 登革热的时空扩散与温度、湿度、居民地、交通、人口密度呈显著相关;登革热病例之间呈显著自相关;登革热发生、扩散与环境、气象、地理、人口中的多种因子存在非线性关系;利用改进的GA-BP神经网络模型模拟登革热时空扩散,均方根误差达到0.081。结论 登革热发生、扩散是由多种因素综合影响的结果;GA-BP神经网络模型能够有效模拟登革热时空过程;此模型同样适用于其他伊蚊类传染病的模拟。     

Coexistence of a cross-diffusive dengue fever model in a heterogeneous environment

A spatio-temporal SIS-SI dengue model with cross-diffusion is formulated as the movements of human and mosquitoes have been intricately linked with the spread of dengue fever. To highlight the impacts of cross-diffusion on the dynamical processes, we focus on the nonnegative steady-state solutions of the dengue model, that is, the coexistence of the corresponding strongly-coupled elliptic system. By means of the relevant eigenvalue problem, we investigate some properties of the basic reproduction number to the model, and further present the existence of coexistence solutions. Our results imply that the virus carried in human and that in mosquitoes can coexist if the basic reproduction number is greater than one and the extent of cross-diffusion is small enough. The final numerical simulations and epidemiological explanations make our analytical findings easier to understand.     

Spatio-temporal modelling of climate-sensitive disease risk: Towards an early warning system for dengue in Brazil

This paper considers the potential for using seasonal climate forecasts in developing an early warning system for dengue fever epidemics in Brazil. In the first instance, a generalised linear model (GLM) is used to select climate and other covariates which are both readily available and prove significant in prediction of confirmed monthly dengue cases based on data collected across the whole of Brazil for the period January 2001 to December 2008 at the microregion level (typically consisting of one large city and several smaller municipalities). The covariates explored include temperature and precipitation data on a 2.5^°×2.5^° longitude-latitude grid with time lags relevant to dengue transmission, an El Niño Southern Oscillation index and other relevant socio-economic and environmental variables. A negative binomial model formulation is adopted in this model selection to allow for extra-Poisson variation (overdispersion) in the observed dengue counts caused by unknown/unobserved confounding factors and possible correlations in these effects in both time and space. Subsequently, the selected global model is refined in the context of the South East region of Brazil, where dengue predominates, by reverting to a Poisson framework and explicitly modelling the overdispersion through a combination of unstructured and spatio-temporal structured random effects. The resulting spatio-temporal hierarchical model (or GLMM—generalised linear mixed model) is implemented via a Bayesian framework using Markov Chain Monte Carlo (MCMC). Dengue predictions are found to be enhanced both spatially and temporally when using the GLMM and the Bayesian framework allows posterior predictive distributions for dengue cases to be derived, which can be useful for developing a dengue alert system. Using this model, we conclude that seasonal climate forecasts could have potential value in helping to predict dengue incidence months in advance of an epidemic in South East Brazil.     

Adaptive Neuro-Fuzzy Inference System for diagnosis risk in dengue patients

Dengue disease is considered as one of the life threatening disease that has no vaccine to reduce its case fatality. In clinical practice the case fatality of dengue disease can be reduced to 1% if the dengue patients are hospitalized and prompt intravenous fluid therapy is administrated. Yet, it has been a great challenge to the physicians to decide whether to hospitalize the dengue patients or not due to the overlapping of the medical diagnosis criteria of the disease. Beside that physicians cannot decide to admit all patients because this will have major impact on health care cost saving due to the huge incident of dengue disease in the country. Even if the physicians managed to identify the critical cases to be hospitalized, most of the tools that have been used for monitoring those patients are invasive. Therefore, this study was conducted to develop a non-invasive accurate diagnostic system that can assist the physicians to diagnose the risk in dengue patients and therefore attain the correct decision. Bioelectrical Impedance Analysis measurements, Symptoms and Signs presented with dengue patients were incorporated with Adaptive Neuro-Fuzzy Inference System (ANFIS) to construct two diagnostic models. The first model was developed by systematically optimizing the initial ANFIS model parameters while the second model was developed by employing the subtractive clustering algorithm to optimize the initial ANFIS model parameters. The results showed that the ANFIS model based on subtractive clustering technique has superior performance compared with the other model. Overall diagnostic accuracy of the proposed system is 86.13% with 87.5% sensitivity and 86.7% specificity.     

Hopfield networks for identification of delay differential equations with an application to dengue fever epidemics in Cuba

This work is aimed at proposing an algorithm, based upon Hopfield networks, for estimating the parameters of delay differential equations. This neural estimator has been successfully applied to models described by Ordinary Differential Equations, whereas its application to systems with delays is a novel contribution. As a case in point, we present a model of dengue fever for the Cuban case, which is defined by a delay differential system. This epidemiological model is built upon the scheme of an SIR (susceptible, infected, recovered) population system, where both delays and time-varying parameters have been included. The latter are thus estimated by the proposed neural algorithm. Additionally, we obtain an expression of the Basic Reproduction Number for our model. Experimental results show the ability of the estimator to deal with systems with delays, providing plausible parameter estimations, which lead to predictions that are coherent with actual epidemiological data. Besides, when the Basic Reproduction Number is computed from the estimated parameter values, results suggest an evolution of the epidemic that is consistent with the observed infection. Hence the estimation could help health authorities to both predict the future trend of the epidemic and assess the efficiency of control measures.     

VIRDOCD: A VIRtual DOCtor to predict dengue fatality

Clinicians make routine diagnosis by scrutinizing patients' medical signs and symptoms, a skill popularly referred to as ‘Clinical Eye’. This skill evolves through trial-and-error and improves with time. The success of the therapeutic regime relies largely on the accuracy of interpretation of such sign-symptoms, analysing which a clinician assesses the severity of the illness. The present study is an attempt to propose a complementary medical front by mathematically modelling the ‘Clinical Eye’ of a VIRtual DOCtor, using statistical and machine intelligence tools (SMI), to analyse Dengue epidemic infected patients (100 case studies with 11 weighted sign-symptoms). The SMI in VIRDOCD reads medical data and translates these into a vector comprising multiple linear regression (MLR) coefficients to predict infection severity grades of dengue patients that clone the clinician's experience-based assessment. Risk managed through ANOVA, the dengue severity grade prediction accuracy from VIRDOCD is found higher (ca 75%) than conventional clinical practice (ca 71.4%, mean accuracy profile assessed by a team of 10 senior consultants). Free of human errors and capable of deciphering even minute differences from almost identical symptoms (to the Clinical eye), VIRDOCD is uniquely individualized in its decision-making ability. The algorithm has been validated against Random Forest classification (RF, ca 63%), another regression-based classifier similar to MLR that can be trained through supervised learning. We find that MLR-based VIRDOCD is superior to RF in predicting the grade of Dengue morbidity. VIRDOCD can be further extended to analyse other epidemic infections, such as COVID-19.     

基于深度森林的高铁站室内热舒适度等级预测

对于高铁站这类半封闭半开放空间的室内环境热舒适度等级难以准确预测的问题,提出基于深度森林（DF）的深度学习方法对热舒适度等级进行科学预测。首先基于现场调研和Energy Plus平台对高铁站室的热交换环境进行建模;其次提炼出客流密度、多联机开行台数和多联机设置温度等8个影响因素,并设计424种工况以获取海量数据;最后采用DF挖掘热舒适度与影响因素之间的关系,以对高铁站室内热舒适度等级进行预测。采用深度神经网络（DNN）和支持向量机（SVM）作为对比算法进行验证。实验结果表明,在3种模型中,DF在预测正确率和weighted-F₁上表现最佳,DF的预测正确率最高达到99.76%,最低为98.11%。因此,DF能够有效预测高铁站室内的热舒适度等级。     

An approximate threshold condition for non-autonomous system: An application to a vector-borne infection

A non-autonomous system is proposed to model the seasonal pattern of dengue fever.We found that an approximate threshold condition for infection persistence describes all possible behaviors of the system.As far as we know, the kind of analysis here proposed is entirely new. No precise mathematical theorems are demonstrated but we give enough numerical evidence to support the conclusions.     

Understanding dengue fever dynamics: a study of seasonality in vector-borne disease models

Dengue fever dynamics show seasonality, with the disease transmission being higher during the warmer seasons. In this paper, we analyse seasonally forced epidemic models with and without vector dynamics. We assume small seasonal effects and obtain approximations for the real response of each state variable and also for the corresponding amplitude and phase via decomposition of the sinusoidal forcing into imaginary exponential functions. The analysis begins with the simplest susceptible-infected-susceptible (SIS) model, followed by the simplest model with vector dynamics, susceptible-infected-susceptible for hosts and uninfected-vector (SISUV). Finally, we compare the more complex susceptible-infected-recovered (SIR) and susceptible-infected-recovered for hosts and uninfected-vector (SIRUV) models and conclude that the models give basically the same information when we replace, in the SIR model, the human infectivity by a function of both human and mosquito infectivities.     

Effect of cross-immunity on the transmission dynamics of two strains of dengue

A deterministic model for the transmission dynamics of two strains of dengue disease is presented. The model, consisting of mutually exclusive epidemiological compartments representing the human and vector dynamics, has a locally asymptotically stable, disease-free equilibrium whenever the maximum of the associated reproduction numbers of the two strains is less than unity. The model can have infinitely many co-existence equilibria if infection with one strain confers complete cross-immunity against the other strain and the associated reproduction number of each strain exceeds unity. On the other hand, if infection with one strain confers partial immunity against the other strain, disease elimination, competitive exclusion or co-existence of the two strains can occur. The effect of seasonality on dengue transmission dynamics is explored using numerical simulations, where it is shown that the oscillation pattern differs between the strains, depending on the degree of the cross-immunity between the strains.     

Targeting the dengue β-OG with serotype-specific alkaloid virtual leads

The dengue envelope β-OG pocket is a crucial hinge for mediating virus-host fusion via conformational changes in the envelope to the fusion-competent form. The β-OG pocket is a small molecule target site for inhibition of virus-host fusion. As of date, the only structure of the β-OG pocket known is of serotype 2. Studies of β-OG inhibition by small molecules primarily target viral serotype 2. Envelope and β-OG sequence alignments, reveal dissimilarities across serotypes. In light of protein sequence-structure-function correlation, sequence variations suggest serotypic variations in β-OG druggability. This, together with the fact that dengue viral proteins do have serotype-specific variations of structure and function, lead to the study of the serotype-specificity of the dengue β-OG ligand binding behaviour. β-OG druggability was compared using comparative models of envelope proteins containing the β-OG pocket in four serotypes of the dengue virus. β-OG ligand binding was found to vary with respect to hydrophobicity, hydrophilicity, hydrogen bonding, van der Waals interactions with ligands and tightness of the binding site. The study also reports serotype-specific virtual leads identified from a library of 9175 alkaloids, using a consensus docking and scoring approach. The docking algorithms of Glide SP and XP, together with the Lamarckian genetic algorithm were employed for consensus docking. For consensus scoring, the Glide empirical score was employed along with the scoring function of AutoDock. A multi-dimensional lead optimisation approach was performed for optimising affinity, ligand efficiency, lipophilic ligand efficiency, ADMET and molecular torsional strains. The study proposes the serotype-specific inhibition of the β-OG for an effective inhibition of virus-host fusion, in contrast to a pan inhibitor.     

Detection of dengue virus serotypes on the surface of gold electrode based on Cratylia mollis lectin affinity

We describe the development of a biosystem to analyze the interactions between CramoLL lectin and fetuin to be applied in the detection of glycoprotein from the serum of patients contaminated with dengue serotypes 1, 2 and 3 (DS1, DS2 and DS3, respectively). The different modified gold electrodes were characterized using cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS) and scanning electron microscopy (SEM) techniques. SEM analyses of PVB-Fe₃O₄Nps-CramoLL-BSA-DS1, PVB-Fe₃O₄Nps-CramoLL-BSA-DS2 and PVB-Fe₃O₄Nps-CramoLL-BSA-DS3 samples have revealed the existence in all cases of a heterogeneous distribution of PVB spherules. EIS and CV measurements have shown that redox probe reactions on the modified gold electrodes were partially blocked, owing to the adsorption of PVB-Fe₃O₄Nps-CramoLL, and confirm the existence of a positive response of the immobilized CramoLL to the presence of fetuin and glycoproteins of dengue serum. The biosensor exhibited a wide linear response to different concentrations of sera of dengue serotypes 1, 2 and 3 and also a higher response to glycoproteins present in dengue serotype 2. These studies might be useful as a platform for the design of new reusable and sensitive biosensing devices that could be of use in dengue diagnosis.     

Spectral prediction of Phytophthora infestans infection on tomatoes using artificial neural network (ANN)

Late blight (LB) is one of the most aggressive tomato diseases in California. Accurately detecting the disease will increase the efficiency of properly controlling the disease infestations to ensure the crop production. In this study, we developed a method to spectrally predict late blight infections on tomatoes based on artificial neural network (ANN). The ANN was designed as a back‐propagation (BP) neural network that used gradient‐descent learning algorithm. Through comparing different network structures, we selected a 3‐25‐9‐1 network structure. Two experimental samples, from field experiments and remotely sensed image data sets, were used to train the ANN to predict healthy and diseased tomato canopies with various infection stages for any given spectral wavelength (µm) intervals. Results of discrete data indicated different levels of disease infestations. The correlation coefficients of prediction values and observed data were 0.99 and 0.82 for field data and remote sensing image data, respectively. In addition, we predicted the field data based on the remote sensing image data and predicted the remote sensing image data with field data using the same network structure, and the results showed that the coefficient of determination was 0.62 and 0.66, respectively. Our study suggested an ANN with back‐propagation training could be used in spectral prediction in the study.     

面向医疗系统的隐私保护疾病预测研究

为了提高医疗数据的隐私性并有效对疾病进行预测,针对从物联网(IoT)设备收集的患者医疗数据,构建了面向医疗系统的隐私保护疾病预测系统框架,通过加密组合文本建立密钥提高了系统认证阶段的隐私性,加强系统和信息传输的安全性。利用基于对数循环值的椭圆曲线密码体制(LR-ECC)提高了数据传输阶段的安全性,从而授权的医护人员可以在医院侧安全地下载患者数据。运用基于象群遗传算法的的深度学习神经网络(EHGA-DLNN)分类技术在疾病预测系统(DPS)阶段实现了疾病数据的有效分类预测。实验结果表明,LR-ECC方法在加密时间和解密时间效率方面高于其他加密方法,并且能够达到98.87%的安全级别,EHGA-DLNN方法在疾病预测分类准确率达到98.35%。     

基于Relief特征选择的心衰死亡率预测

提出心衰死亡率预测系统,预测心衰病人本次住院后30天内死亡率。基于上海曙光医院提供的心衰病人信息,首先对原始数据和特征进行预处理。由于特征的冗余性,再选用经典的Relief特征选择算法筛选出重要的心衰特征,最后选用bp-SVM算法来实现死亡率预测。实验结果证明,死亡率预测系统可以达到较高的性能并通过提供决策信息,辅助医生治疗病人。医生可以根据系统预测的病人死亡率的高低,采取不同的治疗方式,提高临床诊断结果和医院的资源分配。     

基于DFS的专家数据库系统设计

基于DFS,给出了动态模糊专家数据库系统的设计方法,并对系统的功能、系统的结构、DF数据库和规则库进行了详细的描述。     

基于文本挖掘与功能相似性的疾病基因预测

现有疾病基因预测方法大多利用致病基因的各类注释信息进行预测,但仍有很多疾病没有任何注释信息。针对该问题,提出一种基于文本挖掘与功能相似性的疾病基因预测方法,通过数据挖掘获取疾病的相关基因本体术语,利用功能相似性分析基因与疾病之间的相关程度,并根据该相关程度对所有候选基因进行排序,从而识别出致病基因。测试结果显示,该方法能有效预测没有已知功能注释的致病基因。     

基于Hadoop和并行BP网络的打车需求量预测系统研究

一个良好的打车需求量预测系统可以帮助完善城市的交通系统,帮助城市更高效地进行出租车的调度。基于Hadoop设计并搭建了一个打车需求量预测系统。除此之外,针对传统BP神经网络收敛速度慢的缺点,提出了一种基于MapReduce的并行BP神经网络,并将其用作系统的预测模型对打车需求量进行预测。根据实验结果,提出的系统能良好地对城市内某一区域一天内的打车需求量进行预测。     

动态模糊专家数据库系统的设计

本文给出了动态模糊专家数据库系统的设计方法,并对系统的功能,系统的结构,ＤＦ数据库和规则库进行了详细的描述。     

A disease network-based recommender system framework for predictive risk modelling of chronic diseases and their comorbidities

The prediction of chronic diseases and their comorbidities is an essential task in healthcare, aiming to predict patients’ future disease risk based on their previous medical records. The accumulation of administrative data has laid a solid foundation for applying deep learning approaches in healthcare. Existing studies focused on the patients’ characteristics such as gender, age and location to predict the risk of the different diseases. However, there are high dimensional, incomplete and noisy problems in the administrative data. In this research, using administrative health data, we implemented graph theory and content-based recommender system approaches to analyse and predict chronic diseases and their comorbidities. Firstly, we used bipartite graphs to represent the relationships between patients and diseases. Then, we projected this graph to a one-mode graph, namely ‘disease network’. After that, six recommender system models with patient features and network features were trained. The outputs of these models are the severity levels of diseases and the predicted diseases with rank. Finally, we evaluated the performance of these models against the same models without network features. The results demonstrated that the models with network features have lower prediction error and better performances for predicting chronic diseases and their latent comorbidities on large administrative data. Among these models, the graph convolution matrix completion model reveals the least amount of error and the best performance for prediction. Further, using a case study of a specific patient, we demonstrated the application of these models in predictive disease risk analysis. Thus, this study showed the potential application of the recommender system approaches to the health sector utilising administrative claim data, which could significantly contribute to healthcare services and stakeholders.