融入个体活跃度的电子商务客户流失预测模型

为提高个体层次上客户流失预测的精确度, 建立了融入个体活跃度的电子商务客户流失预测模型H-ULSSVM. 该模型首先利用融入地域因素的启发式算法计算出最优阈值, 并求出个体的活跃度, 识别出正判客户和错判客户; 在此基础上, 考虑电子商务客户流失预测影响因素众多, 提出了一种粗糙等价类属性约简方法提取出重要的客户流失预测指标, 然后将降维后的正判客户样本送到非平衡最小二乘支持向量机进行学习和训练, 进而利用得到的分类器对错判客户样本的客户流失状态进行判别. 在某B2C电子商务平台客户样本的实证研究表明, 该模型与其他方法相比, 具有更好的效率和精确度.     

KMG:考虑逆向物流的无人机路径规划策略研究

物流领域无人机派送正成为一种快捷高效的派件方式和应用热点.针对于正向、逆向的物流数据,无人机派送是国内外大型物流企业实施高效物流派送的重要手段.本文提出了一种融合拓展性K-Means++算法和遗传算法的路径动态规划模型(KMG),实现包含逆向物流的无人机调度策略.KMG模型将逆向物流路径融入正向物流路径之中,采用加权聚类算法确定不同属性包裹所需派送无人机的最小数量.在每一簇坐标数据的连通图中,采用遗传算法求解TSP问题,并对可行解进行编码,最终求解出最小欧拉回路.在仿真实验中,KMG模型比独立逆向物流派送的成本减少20.08%,使用拓展性K-Means++聚类计算的时间比传统K-Means算法缩短了298.85%.     

传感器网络中一种能量高效的数据收集算法

为了在无线传感器网络中降低能耗和最大化网络生存期,提出一种能量高效的数据收集算法(energy efficient data gathering algorithm, EEDGA)。该算法利用移动代理模型在网络中进行数据收集。首先,EEDGA根据监测精度的要求控制活动节点的数量;然后,通过求最小支配集得到具体的工作节点;最后,利用蚁群算法规划移动代理迁移的最优路线,移动代理以渐进方式收集活动节点的监测数据。仿真结果表明,与典型算法相比,该算法具有更低的能耗和更长的网络生存期。     

三参数区间数据信息集成算子及其在决策中的应用

研究了三参数区间数据信息的集成问题.基于连续区间数据有序加权平均(C-OWA)算子和有序加权几何(C-OWG)算子,定义了连续三参数区间数据有序加权平均(CP-OWA)算子和有序加权几何(CP-OWG)算子,并将这两种算子进行拓展,提出了加权的CP-OWA(WCP-OWA)算子和加权的CP-OWG(WCP-OWG)算子,研究了它们的一些性质.基于这些算子,提出了一种属性权重和属性值均以三参数区间数形式给出的不确定多属性决策方法,该方法利用CP-OWA算子对三参数区间数属性权重进行处理,利用WCP-OWA算子或WCP-OWG算子对三参数区间数属性值进行集成.最后,进行了实例分析.     

一种区间灰色模糊多准则决策方法

针对准则值和准则权系数为区间灰色模糊数的信息不完全的多准则决策问题,提出了一种灰色模糊多准则决策方法。该方法利用OWA算子对各个方案的准则值进行集结,并对灰色模糊数进行运算,计算各个方案的综合准则值,进而得到各方案的排序。最后,实例说明了该方法的可行性和有效性。     

Clustering routing algorithm of wireless sensor networks based on Bayesian game

To avoid uneven energy consuming in wireless sensor networks, a clustering routing model is proposed based on a Bayesian game. In the model, Harsanyi transformation is introduced to convert a static game of incomplete information to the static game of complete but imperfect information. In addition, the existence of Bayesian nash equilibrium is proved. A clustering routing algorithm is also designed according to the proposed model, both cluster head distribution and residual energy are considered in the design of the algorithm. Simulation results show that the algorithm can balance network load, save energy and prolong network lifetime effectively.     

基于SWEEP方法的改进车辆路径协作策略研究

针对需求随机的车辆路径优化问题,提出了一种基于SWEEP方法的改进车辆路径协作策略,构造了基于该策略的车辆任务量分配模型、设计了求解该模型的启发式算法。该策略采用SWEEP规则对基本车未完成任务的客户重新进行路径优化,然后利用SWEEP车服务这些客户,以缩短客户的服务时间、减少运输成本。应用此方法对24个不同规模的车辆路径优化问题进行了计算机仿真,结果表明,该任务分配模型和算法具有较强的适用性,改进的SWEEP协作策略能够有效地解决解随机车辆路径问题。     

A multi-objective robust operation model for electronic market enabled supply Chain with uncertain demands

A multi-objective robust operation model is proposed in this paper for an electronic market enablexi supply chain consisting of multi-supplier and multi-customer with uncertain demands. Suppliers in this supply chain provide many kinds of products to different customers directly or through electronic market. Uncertain demands are described as a scenario set with certain probability;, the supply chain operation model is constructed by using the robust optimization method based on scenario analyses. The operation model we proposed is a multi-objective programming problem satisfying several conflict objectives, such as meeting the demands of all customers, minimizing the system cost, the availabilities of suppliers＇ capacities not below a certain level, and robustness of decision to uncertain demands. The results of numerical examples proved that the solution of the model is most conservative; however, it can ensure the robustness of the operation of the supply chain effectively.     

基于频繁词集聚类的微博新话题快速发现

新话题发现是进行舆情分析的基础和前提，新话题发现的一个关键环节是进行关键词的聚类分析.目前，大量的新话题来源于微博，但是将传统的聚类算法用于微 博新话题发现时，会产生特征向量的高维性和稀疏性问题，使得聚类结果非常不准确，而且收敛时间难以控制，进而影响舆情分析的可靠性和实时性.鉴此，本文提 出了频繁词集聚类FWSC （frequent words sets clustering）方法.实验结果表明，我们提出的方法能够快速有效地发现新话题.     

基于不确定规划的供应链网络设计模型与算法

利用不确定规划,根据决策者的要求,对供应链网络设计问题进行建模.并采用由随机模拟、模糊模拟以及遗传算法相结合的混合智能算法来求解,最后给出了生活中的实际例子来说明模型和算法的正确性和高效性.     

Design for availability: A holistic approach to create value for manufacturers and customers of capital goods

This research introduces a holistic framework called Design for Availability that uses the principles of Lean Sigma and Design for X to cost-effectively optimize the availability of capital goods(i.e., technical systems used in the production of end-products or -services such as medical systems, airplanes,and manufacturing equipment) throughout their entire lifetime.Manufacturers require such a framework because users of capital goods increasingly insist on high system availability levels against reduced lifetime costs.The Design for Availability framework allows manufacturers to determine the current status of system availability and associated lifetime costs,and to identify opportunities to create additional value for themselves and their customers.A case study at a global manufacturer of capital goods in the food processing industry illustrates how the framework can be used in practice and to what extent the manufacturer and customers may profit from applying Design for Availability.     

传感器网络中一种基于时分的路由算法

针对采用多跳传输数据时靠近基站的节点易耗尽能量而失效、单跳传输数据时距离基站较远的节点很快失效的问题,提出了"2跳"的概念和基于时分的路由算法,其基本思想是在不同时间段分别采用单跳、多跳和"2跳"路由。仿真实验表明,算法使节点能量消耗比较均匀,能延长网络的生命周期。     

Induced and heavy aggregation operators with distance measures

This paper introduces a new aggregation model by using induced and heavy aggregation operators in distances measures such as the Hamming distance. It is called the induced heavy ordered weighted averaging (OWA) distance (IHOWAD) operator. This paper studies some of its main properties and a wide range of particular cases such as the induced heavy OWA (IHOWA) operator, the induced OWA distance (IOWAD) operator and the heavy OWA distance (HOWAD) operator. This approach is generalized by using generalized and quasi-arithmetic means obtaining the induced generalized IHOWAD (IGHOWAD) operator and the Quasi-IHOWAD operator. An application of the new approach in a decision making problem regarding the selection of strategies is developed.     

Applying particle swarm optimization and honey bee mating optimization in developing an intelligent market segmentation system

Data mining has been successfully applied in many fields to find useful information stored in vast databases.Market segmentation,which segments data into homogenous clusters by using cluster analysis,is among the most important of the applications in data mining.In this study,we propose a clustering system which integrates particle swarm optimization and honey bee mating optimization methods (PSHBMO).Simulations for a benchmark test function show that our proposed method is better equipped to find the global optimum than other well-known clustering algorithms.Finally,the proposed clustering system is applied to a real-world consumer electronic company to perform market segmentation via the RFM model.     

虚拟氛围对在线消费者冲动购买意向影响的实证研究

根据环境心理学理论,个体行为与其所处的环境是相互影响的。以环境心理学中的Mehrabian-Russel模型为基础,结合沉浸理论,研究了在线环境下虚拟氛围的各种要素对消费者冲动购买意向的影响。实证研究以购买决策涉入度作为控制变量,分析了电子商务系统中的虚拟氛围要素(导航易用性、信息性、社会性与娱乐性)与在线消费者沉浸状态要素(愉悦性感知、行为控制感知与专注)的影响关系,进而发现了虚拟氛围要素、消费者沉浸状态要素影响在线消费者冲动购买意向的机理。     

A Fuzzy Negotiation Model of E-commerce

We usually decide to accept or reject the offer based on the utility function in the traditional automatic negotiation of E-commerce so that we reject the offers whose utility is lower than a specified value. Here we evaluate the acceptability based on the fuzzy set theory and the membership function. Since different issues have different effect on the negotiators, we state the combined concession in the multi-issue negotiation for the negotiators. We put forward a more practical negotiation model than the traditional negotiation model and give an example of its application in E-commerce.     

考虑客户时间偏好的第四方物流路径优化问题

为使第四方物流系统能够在不确定环境下为客户提供有效的运输方案,在一定费用投入下获得更高的客户满意度,研究考虑客户时间偏好的第四方物流路径优化问题.基于累积前景理论,以最大化总运输时间的前景值为目标,建立考虑客户时间偏好的数学模型,并采用蚁群算法对模型进行求解,数值算例验证了算法的有效性·并且,将该模型与传统的期望值模型和期望效用模型进行对比,算例分析表明,考虑客户时间偏好模型可以更有效地描述客户心理行为,并适用于具有不同风险态度的客户群体,验证了模型的有效性.     

单周期离散随机需求的库存-运输整合优化

库存-运输整合优化问题极为复杂,目前尚缺乏系统研究。本文针对一个配送中心与多个客户组成的配送系统,讨论客户单周期离散随机需求条件下,建立该系统的库存-运输整合优化问题模型。并通过分析模型的离散特征,把问题转化为一般的车辆路径问题,借鉴现有的求解方法,构造了求解模型的自适应单亲遗传算法。最后用MATLAB编程并用测例进行了测试。结果表明,借助现代优化方法,库存-运输整合问题的研究具有理论意义和实际价值。     

基于Choquet积分的HMM商品信息抽取方法

电子商务网站使用不同的网页编码技术和页面布局为比较购物信息获取带来了很大的难度.基于隐马尔可夫模型(HMM)的信息抽取模型有着易于建立、适应性强等优点,被视为一种有效的信息抽取方法.但是这种算法存在状态序列计算复杂、难以训练优化抽取模型等缺点.本文应用模糊积分单调性建立基于Choquet积分的隐马尔可夫模型(CI-HMM),解决HMM观察序列概率计算所需的条件独立性假设,优化HMM观察序列的计算.本文以网上书店商品数据进行实证,实验证明CI-HMM比HMM有更好适用性和精确度.     

拓展的C2OWA 算子及其在不确定多属性决策中的应用

把Yager提出的连续区间数据OWA(C-OWA)算子进行拓展,提出了加权的C-OWA(WC-OWA)算子、有序加权的C-OWA(OWC-OWA)算子、以及组合的C-OWA(CC-OWA)算子,研究了它们的一些性质.基于这些算子,分别在单人决策和群决策这两种情形下,提出了属性权重确知、且属性值以区间数形式给出的不确定多属性决策方法.最后,进行了实例分析.