基于Logit离散选择模型的品类优化问题综述

品类优化问题(Assortment Optimization Problem)是收益管理的经典问题.它研究零售商在满足运营约束的前提下,应如何从给定产品集合中选择一个子集提供给消费者,以最大化预期收益.该问题的核心在于如何准确地刻画消费者在面对细分产品时的选择行为、建立相应的优化模型并设计高效率的求解算法.基于Logit离散选择模型的品类优化问题:首先,介绍了基于Multinomial Logit模型的品类优化问题.然后介绍了两个更复杂的变种:第一个是基于两层以及多层Nested Logit模型的品类优化问题,这类问题可合理刻画细分产品之间的"替代效应";第二个是基于Mixtures of Multinomial Logits模型的品类优化问题,这类问题可充分考虑消费者群体的异质性.随后,介绍了数据驱动的品类优化问题的相关进展.最后,指出该问题未来可能的若干研究方向.     

基于位移型控制方程的圆板热后屈曲分析

本文推导了圆板位移型热后屈曲方程，在设定挠度试函数后，由微分方程精确求出径向位移，然后用Galerkin法消去另一方程的残差，试函数由Legendre多项式构成，结果表明：本文方法是有效的，有关结果可供设计圆板时参考。     

基于信息熵的社交网络观点演化模型

随着网络服务的发展,社交网络逐渐成为信息传播的新媒介.因此,研究网络舆情演化具有重要意义和实用价值.为了更好地研究网络舆论,在信息熵的基础上,提出了一个社交网络观点演化模型.此模型存在以下两个特点:一是可以反映个体面对正负两种观点趋向做出抉择时的心理过程;二是可以反映个体形成新观点时主观因素和客观因素的影响.在仿真实验中,讨论了舆论环境对个体观点演化的影响,初始观点和自信度对观点演化的影响,以及意见领袖对群体观点演化的影响.实验结果表明,该模型可以反映真实社交网络中个体的心理学特征,比如个体的观点形成会受到舆论环境的影响,自信的个体不愿意接受他人的观点,当意见领袖存在时群体的观点会受到影响等.     

基于级联模型的中文情感要素抽取

随着社交媒体的发展及成熟,每天在互联网环境中都会产生大量的用户评论信息。抽取评价短语、评价对象和观点持有者等情感要素,已经成为了中文观点挖掘和情感分析的重要先决任务。针对中文情感要素抽取任务,本文提出了一个统计和规则相结合的级联模型,主要贡献包括：（1）针对汽车领域评论信息,构建情感要素标注语料库和相关词典;（2）对于以往研究较少关注的中文评价短语,本文详细分析阐述其定义和分类;（3）结合统计和规则,分别针对评价短语和情感要素提出级联抽取策略。实验结果充分证明了该级联模型的有效性,相比较于其它基于规则的情感要素抽取算法有效提升了召回率,同时为后续社交媒体情感分析任务提供了有力的支持。     

Social influence and opinions

In this paper we describe an approach to the relationship between a network of interpersonal influences and the content of individuals’ opinions. Our work starts with the specification of social process rather than social equilibrium. Several models of social influence that have appeared in the literature are derived as special cases of the approach. Some implications for theories on social conflict and conformity also are developed in this paper.     

Scale-Free Interpersonal Influences on Opinions in Complex Systems

An important side effect of the evolution of the human brain is an increased capacity to form opinions in a very large domain of issues, which become points of aggressive interpersonal disputes. Remarkably, such disputes are often no less vigorous on small differences of opinion than large differences. Opinion differences that may be measured on the real number line may not directly correspond to the subjective importance of an issue and extent of resistance to opinion change. This is a hard problem for the field of opinion dynamics, a field that has become increasingly prominent as it has attracted more contributions to it from investigators in the natural and engineering sciences. The paper contributes a scale-free approach to assessing the extents to which individuals, with unknown heterogeneous resistances to influence, have been influenced by the opinions of others.     

Individual decision‐making in dynamic collective systems

One of the most popular individual choice models is the multinomial logit model (MLM). The basic hypothesis of the MLM is that an individual chooses an alternative among a set of alternatives available to him by comparing the utilities of all the choices. The MLM is, however, limited in its usefulness because it doesn't account for the effects of the time factor and the social interactions between individuals. This work describes a dynamic extension of the MLM that allows for such interactions. The new model is formulated as an interactive continuous‐time Markov process, and is approximated, for a large population, by a deterministic system. Some possible consequences of the interaction phenomena are discussed with a special binary choice model.     

Optimal control of the freezing time in the Hegselmann–Krause dynamics

We study the optimal control problem of minimizing the freezing time in the discrete Hegselmann–Krause (HK) model of opinion dynamics. The underlying model is extended with a set of strategic agents that can freely place their opinion at every time step. Indeed, if suitably coordinated, the strategic agents can significantly lower the freezing time of an instance of the HK model. We give several lower and upper worst-case bounds for the freezing time of a HK system with a given number of strategic agents, while still leaving some gaps for future research.     

Oscillation of the NCO model

We study opinion oscillation of the nonconsensus opinion model (NCO) on graphs, and in particular on bipartite graphs. Using intensive numerical simulations, we investigate the relationship between amplitude A$A$ (the percentage of nodes whose opinions oscillate) and (p,q)$(p,q)$, which are the initial configuration fractions with opinion 1 on two sets of two bipartite graphs. Finally, for the general graph, we present several definitions and develop three propositions as regards whether an oscillation can occur or not on a certain graph.     

短信舆情信息处理平台及关键技术

分析了短信舆情信息处理平台的结构设计和系统处理流程等方面的问题,开发了完整短信舆情监管系统,介绍了平台中的短信信息采集模块、短信信息处理模块、热点信息发现模块的主要功能,提出了一种适合短信数据的数据采集方法及热点发现方法。系统运行结果表明,系统能有效保证短信网络内容安全,准确率较高。