用神经网络组合预测算卫生总费

利用BP神经网络组合能够较好地模拟在各种不确定因素影响下因果变量之间的内在关系。建立了基于人工神经网络的卫生总费用预测模型,该模型的网络结构由输入层(1个节点)、隐层(7个节点)和输出层(1个节点)组成。采用改进的BP算法对7组学习样本进行训练,得到各节点间的连接权和阈值,然后用优化好的网络进行卫生总费用预测。预测结果表明,利用该方法建立的模型预测误差在1%以内,十分精确。     

用神经网络组合预测算卫生总费用

利用BP神经网络组合能够较好地模拟在各种不确定因素影响下因果变量之间的内在关系.建立了基于人工神经网络的卫生总费用预测模型,该模型的网络结构由输入层(1个节点)、隐层(7个节点)和输出层(1个节点)组成.采用改进的BP算法对7组学习样本进行训练,得到各节点间的连接权和阈值,然后用优化好的网络进行卫生总费用预测.预测结果表明,利用该方法建立的模型预测误差在1%以内,十分精确.     

灰色神经网络在股价预测中的应用

本文在深入了解股价的可预测性和研究各种股价预测的方法的基础上,探讨利用BP神经网络和灰色模型进行股价预测。以二者为基础,将BP神经网络和灰色模型预测模型的优点结合,提出灰色神经网络模型。该模型将灰色模型的预测值作为BP神经网络的输入变量,而以实际值作为输出变量对神经网络进行训练。     

基于BP神经网络的中国制造业生产率预测模型——交叉学科在生产物流领域的应用研究

BP网络能学习和存贮大量的输入-输出模式映射关系,而无需事前揭示描述这种映射关系的数学方程。因此采用三层BP网络实现制造业工业增加值、制造业全社会固定资产投资和工资向全要素生产率的转换,借助MATLAB神经网络工具箱编写了训练程序、测试程序、预测程序,最终神经网络隐层含有13个节点,传递函数采用tansig函数;输出层传递函数选用purelin函数,得到的训练误差为8.44272×10-6,结果满意,可以认为该神经网络可以用来实现这个关系映射,并对2007年全要素生产率进行了预测。     

基于BP的CRM系统销售预测的数据挖掘

提出了一种基于改进BP神经网络的销售预测模型.该模型通过数据仓库获取销售历史数据,利用BP网络进行走势预测.系统采用VC++实现,取得成功.     

基于遗传算法优化混沌神经网络的股票指数预测

为提高BP神经网络预测模型对混沌时间序列的预测准确性,提出一种基于遗传算法优化BP神经网络的改进混沌时间序列预测方法。本文采用时间序列输入输出参数数量构造BP神经网络拓扑结构,利用遗传算法优化BP神经网络的权值和阈值,然后训练BP神经网络预测模型求得最优解,将该预测方法应用到上证综合指数的时间序列进行有效性验证,结果表明了该方法对上证综合指数具有更好的非线性拟合能力和更高的预测准确性。     

广义回归神经网络的金融预测模型研究

本文针对BP网络存在着收敛速度慢和局部极小的问题,提出了一种基于广义回归神经网络方法的金融预测模型,该网络运用于汇率模拟与预测,以演示训练样本的构建、原始数据预处理、神经网络的创建训练和检测结果的评价整个过程.通过详细的仿真实验以及与BP神经网络的比较可得出以下结论:该方法不仅运算速度较快,且逼近性能及预测性能明显都优于传统BP神经网络.     

BP神经网络的C2C电子商务信用评价模型的建立

从店铺信息、商品、卖家服务等方面建立C2C电商信用评价体系,在淘宝网获取了242条卖家信息,利用SPSS两步聚类法将数据进行聚类,运用MATLAB的BP神经网络工具箱,将评价指标数值作为输入,聚类结果作为输出,构建了三层BP神经网络,以230条数据为训练样本,得到具有较快收敛速度和较高准确率的BP网络。在此基础上,以12条卖家信息作仿真实验,对卖家信用等级进行客观的预测评价,最终建立C2C信用评价模型。     

基于人工神经网络的煤炭需求预测

煤炭是人类赖以生存的能源物质之一,有效地预测煤炭能源需求对于社会的发展有重要意义。BP神经网络预测模型具有自学习、自适应的特点,适合用于难于建立精确数学模型的系统。本文综合考虑影响煤炭需求的各个因素,通过改进的BP模型进行煤炭需求预测,并用MATLAB仿真实现,该预测结果有很好的适用价值。     

基于BP神经网络的中小企业生产运作能力评价研究

为了提高中小企业生产运作能力,建立了中小企业生产运作能力评价指标体系,并利用BP神经网络建立了中小企业生产运作能力的评价模型。为减小希望输出与实际输出之间的误差,从输出层神经元到输入层神经元逐层反向修正各连接权,并在网络训练学习的过程中不断执行误差反向传播修正,以提高网络对输入模式响应的正确率。     

New Technology in Shopping: Forecasting Electronic Shopping With the Use of a Neuro-Fuzzy System

This article presents the application of neuro-fuzzy techniques in forecasting a new technology in shopping. Neural networks have been used successfully to forecast time series due to their significant properties of treating nonlinear data with self-learning capability. However, neural networks suffer the difficulty of dealing with qualitative information and the “black box” syndrome that more or less limits their applications in practice. To overcome the drawbacks of neural networks, in this study, we proposed a fuzzy neural network that is a class of adaptive networks functionally equivalent to a fuzzy inference system. The results derived from the experiment based on electronic sales indicated that the suggested fuzzy neural network could be an efficient system to forecast a new technology in shopping. Experimental results also show that the neuro-fuzzy approach outperforms the other two conventional models (AR and ARMA).     

Network structure,equilibrium and dynamics in a monopolistically competitive economy

Although network theory has been busy to emphasize the role of connection structures in shaping aggregate level phenomena of complex systems, there are only few attempts in economic modeling which try to build this dimension into the analysis. Macroeconomic models typically build on complete connectedness among economic actors (frictionless flow of information, perfect information on prices), thus these models typically oversee the possible effects of complex, incomplete network structures among economic agents on emergent macroeconomic phenomena. In this paper we try to fill this gap by incorporating possibly incomplete relationship structures between economic actors in a standard model of monopolistic competition and then analyze the effect of different network structures on the equilibrium and dynamic properties of the model. Analytical and simulation results of the model show that incomplete connectedness give rise to deadweight loss, shrinking output below the level observed in standard models with complete networks. Also, the dynamics of link formation has an effect on the steady state of the economy as well as on its response to shocks.     

Predicting total health care costs of medicaid recipients: An artificial neural systems approach

Although medical treatment costs have escalated beyond the reach of many Americans, a thorough total cost model is essential before implementing cost containment strategies. This study offers a prediction model of the total treatment cost for a Mississippi Medicaid patient. Artificial neural systems (ANS) are proposed as a methodology for the prediction of health care costs of postmenopausal women who are Medicaid recipients. The results of the neural networks along with traditional regression analysis are presented. Artificial neural systems overcome many of the problems associated with the estimation of this model, such as the identification of the appropriate functional form and dealing with both qualitative and quantitative aspects of these large claims databases. Neural networks are shown to provide superior forecasts. In addition preliminary results for the presentation of significance tests of individual causal variables using neural networks is presented.     

经济增长不确定性的自组织机制分析

如何阐释和说明世界各国经济发展路径的差异性、增长状态的多样性,建立经济增长不确定性理论,是90年代末以来国外经济增长研究的一个重要分支。本文从经济系统的自组织性出发,基于经济增长的要素性质、多部门组成,应用系统自组织理论,研究经济增长的路径演化、状态转移和结构变迁,以建立经济增长不确定性的自组织机制。     

To post or not to post: social media sharing and sporting event performance

Social networks are becoming increasingly important for consumers, especially in the context of sport, where the service experience is highly intense. Few studies have combined subjective event performance variables and social network variables to analyze social network content sharing by sports practitioners. This article investigates the use of social networks in relation to sporting events. An empirical study examined the role of social network variables and sporting event performance variables in social media use. The sample consisted of 410 triathletes (72.2% male) aged between 18 and 66 years (mean 37.03 ± 8.62). Four analyses were performed using fuzzy‐set qualitative comparative analysis to examine the causes of sharing comments through social media, sharing photos and videos on social media, participant satisfaction, and word‐of‐mouth (WOM). The event's general image was a necessary condition in all cases. The combination of participants’ satisfaction and positive event image and the combination of social network use and positive event image lead to social network content sharing by athletes. The combination of positive event image and participant satisfaction leads to a positive WOM.     

国家知识产权战略示范省评价指标体系研究

知识产权作为21世纪新的经济资源,成为各国综合国力提升的原动力,成为各国综合国力的核心竞争力。我国2008年颁布国家知识产权战略纲要,继而授予江苏省国家知识产权战略示范省。本文基于知识产权和知识产权战略的基础理论,借鉴国家、上海等知识产权战略评价指标体系建构原则和经验的基础上,建立了系统科学的创建实施国家知识产权战略示范省工程的评价指标体系,并对主要指标进行了解释。     

移动传感器网络的安全性能智能预测

移动传感器网络的物理层安全问题日益复杂,已经成为了一个研究热点。为了及时处理网络安全事件,研究了移动传感器网络的安全性能预测,提出了一种基于灰狼优化广义回归（Grey Wolf Optimization-Generalized Regression,GWO-GR）神经网络的安全性能智能预测方法。该方法利用发射天线选择策略,推导了非零安全容量概率性能的精确闭合表达式。仿真比较了所提方法、反向传播神经网络、广义回归神经网络、支持向量机等方法,结果表明,所提方法可以实现更好的预测性能,提高安全性能预测的实时性。     

基于深度学习的远程视频监控异常图像检测

针对复杂场景下远程视频监控图像异常检测困难、传统算法功能单一(仅针对某种特定场景或某种异常图像进行检测)等问题,提出一种基于深度学习的全自动远程视频异常图像检测方法。首先采用Xavier方法对自行设计的卷积神经网络(Convolutional Neural Network,CNN)的参数进行初始化,然后将标准化后的视频差分图送入CNN的输入层,通过特征提取及下采样,最后在CNN的输出层获得远程视频异常图像检测结果。实验结果表明,该方法可以对远程视频监控中突然出现遮挡、模糊和场景切换等多种异常同时进行实时在线检测,准确率可达88.75〖WT《Times New Roman》〗%〖WTBZ〗。