带有能力约束的大规模联合补货问题研究

文中研究求解带有能力约束限制的大规模联合补货问题。在订货网络中,通过聚类算法将问题规模缩小至一定范围内,使算法在该范围内具有良好的表现;其次,在问题的模型中,考虑转运中心的能力限制,保证在任何订货周期内不出现超出转运中心能力负荷的情况;第三,通过改进经典RAND算法,引入拉格朗日乘子,求解模型,所优化的目标为总费用最小化和求解速度质量的平衡。最后根据数值算例验证策略有效性。     

基于聚类—神经网络的客户价值分类方法及应用

本文从价值客户管理角度出发,建立了一种基于聚类——神经网络的客户价值分类模型。阐述了客户价值分析与评价指标设计、分类建模原理与网络学习算法选择、网络学习的前导原理与聚类算法、建模流程设计,并以某地区小灵通客户价值分类预测模型的使用及客户管理策略为应用实例。提出了研究的框架与方法,为企业客户价值管理决策系统的开发提供了重要模型依据。     

对外贸易对我国碳排放的影响分析

能源是人类社会赖以生存和发展的重要物质基础,是促进经济增长最基本的要素之一。随着世界经济的迅速发展,气候变化对人类活动的影响越来越深,也更加受到国际社会的重视。中国正处于工业化和城市化快速发展的过程,经济增长与资源供给和环境保护的矛盾日益尖锐,特别是加入世贸组织以来,经济高速发展,对外贸易不断扩大,随之而来的环境问题越来越受到人们的关注。本文应用聚类分析的方法,得出对外贸易与碳排放量之间关系以加入世贸组织为标志分为两类,并应用回归分析对比入世前后对外贸易对我国碳排放量的影响变化,结果表明：入世后,国内碳排放量与对外贸易的弹性值变大,说明我国对外贸易的增长伴随着较高的能耗水平,政府在协调环境保护与经济发展的关系上仍需努力。     

企业“扎堆”、技术升级与经济绩效——开发区集聚效应的实证分析

开发区是中国经济转型过程产业空间组织的一种"转型制度"形式,正在经历一个企业"扎堆"向产业集群的转变。本文依据江苏省沿江开发区企业的微观调研数据,检验以开发区为载体的产业空间集聚因素与技术升级对"落地"企业经济绩效的影响,从而揭示以开发区为载体的企业集聚的性质以及演变为产业集群的趋势。实证研究表明,企业进驻开发区的主要目的是获取"政策租",由此导致开发区企业"扎堆"不具有一般意义上的产业集聚效应:靠近要素或者靠近市场的好处。企业一旦"落地"开发区,尽管它并不必然与关联企业在空间上集中,但其技术和市场行为必将与当地政府和产业关联要素发生作用,从而影响企业的经济绩效和开发区的集聚效应。因此,在产业转移和国家经济政策调整导致"政策租"耗散的情况下,基于开发区的企业集群可持续发展的关键在于提升开发区的非政策性的集聚效应。     

1990-2004中国城市体系演变研究

通过考察人口在各级别城市的分布,利用住序-规模法则验证,对各省城市体系类型聚类分析,考察城市体系等级规模的变化,分析人口经济总量的分布现状及其演变,发现我国城市化的总体趋势是城市化进程比较快,呈现出特大城市和中小城市优先发展的事实.城市体系演变的趋势是城市体系由低级向高级、低水平向高水平、由不平衡向平衡型的演变.整体看我国城市的发展势头良好,但无论从绝对差距还是从相对差距看,东部地区相对快速发展的趋势还会持续下去.我国城市等级规模及人口经济规模分布极不均衡,在一定程度上影响区域经济的发展,应当分区域采取不同的发展战略.今后国家城市建设应当集约利用土地,重点扩大城市人口规模.国家在投资鼓励、区域发展政策等方面应当对中西部地区有实质性的扶持,努力使中西部地区城市人口与经济实力增强.     

A dual-clustering algorithm for a robust medical grid partition problem considering patient referral

The medical grid including hospitals at all levels is a new hierarchical diagnosis and treatment system. It is assumed to provide health services for residents in a certain area, allowing free referral of patients, so as to effectively utilize medical resources. Therefore, from the perspective of the government, the key issue is how to divide the medical grid in a robust and balanced manner. In this paper, various deterministic factors, such as hospital level, location and department, as well as uncertain factors, including patient distribution or population density, are considered in decision-making. To solve this problem, a dual-clustering algorithm based on K-means and K-medoids (DCKK) is developed with local search methods to minimize the average patient waiting and travelling time. The experimental results show that DCKK algorithm can generate better and more robust grid partition solutions than the existing mainstream algorithms in different scenarios. In addition, the rules between the number of medical grids and the number of patients, as well as the hospital sharing between medical grids, are also studied. Finally, a real medical grid partition case of Ji'nan, China, with forty hospitals in four urban areas, is studied, and five medical grids are recommended.     

Measurement of development and regional disparities in Greek periphery: A multivariate approach

The identification of regional disparities and regional growth patterns is an important factor affecting policy formulation. Single indicator, usually GDP-based, approaches have revealed significant shortcomings. In this work we provide a methodology and the respective tools to analyse regional disparities and development patterns. Aiming at capturing the different aspects of development and quality of life our approach is multi-dimensional: we, first, develop a composite index; we, then, apply multivariate clustering for identifying regions with similar socio-economic profiles. The methodology is applied to examine Greek regions. The results do not provide strong evidence for convergence of Greek regions during the period 1995–2007.     

我国能源生产率的地区划分及影响因素分析

能源生产率状况以其对能源消费量的重要影响而引起了国内外理论界的广泛关注。本文整理了我国各个省区1995～2003年的数据,并以此为基础来研究这些省区能源生产率的分类情况及差别原因。通过聚类分析,本文将各省划分为能源高效区、中效区和低效区三类,并利用面板数据计量分析了经济发展水平、产业结构、投资情况以及能源价格等因素对能源生产率的影响。     

Self-organization of R&D search in complex technology spaces

We extend an earlier model of innovation dynamics based on percolation by adding endogenous R&D search by economically motivated firms. The {0, 1} seeding of the technology lattice is now replaced by draws from a lognormal distribution for technology ‘difficulty’. Firms are rewarded for successful innovations by increases in their R&D budget. We compare two regimes. In the first, firms are fixed in a region of technology space. In the second, they can change their location by myopically comparing progress in their local neighborhoods and probabilistically moving to the region with the highest recent progress. We call this the moving or self-organizational regime (SO). The SO regime always outperforms the fixed one, but its performance is a complex function of the ‘rationality’ of firm search (in terms of search radius and speed of movement). The clustering of firms in the SO regime grows rapidly and then fluctuates in a complex way around a high value that increases with the search radius. We also investigate the size distributions of the innovations generated in each regime. In the fixed one, the distribution is approximately lognormal and certainly not fat tailed. In the SO regime, the distributions are radically different. They are much more highly right skewed and show scaling over at least 2 decades with a slope around one, for a wide range of parameter settings. Thus we argue that firm self-organization leads to self-organized criticality. The online version of the original article can be found under doi:.