Comparing supervised learning algorithms and artificial neural networks for conflict prediction: performance and applicability of deep learning in the field

Quality & Quantity - Machine learning algorithms and artificial neural networks promise a new and powerful approach for making better and more transferable predictions in global conflict...     

Spatial social networks

We introduce a spatial cost topology in the network formation model analyzed by Jackson and Wolinsky, Journal of Economic Theory (1996), 71: 44–74. This cost topology might represent geographical, social, or individual differences. It describes variable costs of establishing social network connections. Participants form links based on a cost-benefit analysis. We examine the pairwise stable networks within this spatial environment. Incentives vary enough to show a rich pattern of emerging behavior. We also investigate the subgame perfect implementation of pairwise stable and efficient networks. We construct a multistage extensive form game that describes the formation of links in our spatial environment. Finally, we identify the conditions under which the subgame perfect Nash equilibria of these network formation games are stable. We are very grateful for the constructive comments of Matt Jackson and an anonymous referee. We also like to thank Vince Crawford, Marco Slikker, Edward Droste, Hans Haller, Dimitrios Diamantaras, and Sudipta Sarangi for comments on previous drafts of this paper.We acknowledge Jay Hogan for his programming support. Part of this research was done while visiting the CentER for Economic Research, Tilburg University, Tilburg, The Netherlands.Financial support from the Netherlands Organization for Scientific Resrarch (NWO), grant B46-390, is gratefully acknowledged.-->,     

Remaining life estimation of used components in consumer products: Life cycle data analysis by Weibull and artificial neural networks

Environmental awareness and legislative pressures have made manufacturers responsible for the take-back and end-of-life treatment of their products. To competitively exploit these products, one option is to incorporate used components in “new” or remanufactured products. However, this option is partly limited by a firm's ability to assess the reliability of used components. A comprehensive two-step approach is proposed. The first stage phase statistically analyzes the behavior of components for reuse. A well-known reliability assessment method, the Weibull analysis, is applied to the time-to-failure data to assess the mean life of components. In the second phase, the degradation and condition monitoring data are analyzed by developing an artificial neural network (ANN) model. The advantages of this approach over traditional approaches employing multiple regression analysis are highlighted with empirical data from a consumer product. Finally, the Weibull analysis and the ANN model are then integrated to assess the remaining useful life of components for reuse. This is a critical advance in sustainable management of supply chains since it allows for a better understanding of not only service requirements of product, but the remaining life in a product and hence its suitability for reuse or remanufacture. Future work should assess: (1) reduction in downtime of process equipment through the implementation of this technique as a means to better manage preventative maintenance; (2) reduce field failure of remanufactured product; (3) selling-service strategy through implementation of the proposed methodology.     

Predicting criminal recidivism using neural networks

Prediction of criminal recidivism has been extensively studied in criminology with a variety of statistical models. This article proposes the use of neural network (NN) models to address the problem of splitting the population into two groups — non-recidivists and eventual recidivists — based on a set of predictor variables. The results from an empirical study of the classification capabilities of NN on a well-known recidivism data set are presented and discussed in comparison with logistic regression. Analysis indicates that NN models are competitive with, and may offer some advantages over, traditional statistical models in this domain.     

Information flows and crashes in dynamic social networks

We develop a dynamic model of information transmission and aggregation in social networks in which continued membership in the network is contingent on the accuracy of opinions. Agents have opinions about a state of the world and form links to others in a directed fashion probabilistically. Agents update their opinions by averaging those of their connections, weighted by how long their connections have been in the system. Agents survive or die based on how far their opinions are from the true state. In contrast to the results in the extant literature on DeGroot learning, we show through simulations that for some parameterizations the model cycles stochastically between periods of high connectivity, in which agents arrive at a consensus opinion close to the state, and periods of low connectivity, in which agents’ opinions are widely dispersed.

     

Forecasting exchange rates using feedforward and recurrent neural networks

In this paper we investigate the out-of-sample forecasting ability of feedforward and recurrent neural networks based on empirical foreign exchange rate data. A two-step procedure is proposed to construct suitable networks, in which networks are selected based on the predictive stochastic complexity (PSC) criterion, and the selected networks are estimated using both recursive Newton algorithms and the method of nonlinear least squares. Our results show that PSC is a sensible criterion for selecting networks and for certain exchange rate series, some selected network models have significant market timing ability and/or significantly lower out-of-sample mean squared prediction error relative to the random walk model.     

Identification of peer effects through social networks

We provide new results regarding the identification of peer effects. We consider an extended version of the linear-in-means model where interactions are structured through a social network. We assume that correlated unobservables are either absent, or treated as network fixed effects. We provide easy-to-check necessary and sufficient conditions for identification. We show that endogenous and exogenous effects are generally identified under network interaction, although identification may fail for some particular structures. We use data from the Add Health survey to provide an empirical application of our results on the consumption of recreational services (e.g., participation in artistic, sports and social activities) by secondary school students. Monte Carlo simulations calibrated on this application provide an analysis of the effects of some crucial characteristics of a network (i.e., density, intransitivity) on the estimates of peer effects. Our approach generalizes a number of previous results due to Manski [Manski, C., 1993. Identification of endogenous social effects: The reflection problem. Review of Economic Studies 60 (3), 531–542], Moffitt [Moffitt, R., 2001. Policy interventions low-level equilibria, and social interactions. In: Durlauf, Steven, Young, Peyton (Eds.), Social Dynamics. MIT Press] and Lee [Lee, L.F., 2007. Identification and estimation of econometric models with group interactions, contextual factors and fixed effects. Journal of Econometrics 140 (2), 333–374].     

Neighborhood social networks and female self-employment earnings in Taiwan

Using the sample of Taiwanese female self-employed workers from the Taiwan Women and Family Survey (TWFS) conducted in 1989, this article explores neighborhood social networks and female self-employment earnings. The evidence from linear regression and endogenous treatment effect models suggests that an individual’s past provision of personalized assistance to adjacent neighbors significantly increases current self-employment earnings, with corrections for self-selection into network participation. An individual’s past voluntary contributions to community organizational activities also significantly increase current earnings, both with and without corrections for self-selection. However, there are negative self-selections into both cooperative neighborhood networks and community organizational activities in the sense that female self-employed workers with high earnings capacity are less likely to choose to participate in these types of social networks.     

Influences on expatriate social networks in China

This study applied social network theory to investigate the factors influencing expatriate social networks and the consequences of expatriate social networks in China. Based on analysis of 171 Taiwanese expatriates in China, this study found that core self-evaluations and extraversion are significant for expatriates in developing expressive and instrumental ties with host country nationals. Moreover, job autonomy assists expatriates in developing instrumental ties with host country nationals. The expressive and instrumental ties of expatriates with host country nationals are significant for overseas adaptation. Finally, instrumental ties with host country nationals are significant for expatriate job performance.     

Research collaboration model in academic social networks

ABSTRACT

This paper proposes a hybrid collaboration recommendation method that accounts for research similarities and the previous research cooperation network. Research cooperation is measured by combining the collaboration time and the number of co-authors who already collaborated with at least one scientist. Research similarity is based on authors’ previous publications and academic events they attended. A weighted directed graph is built to discover new collaborators by using direct and indirect connections between scientists. Moreover, a consensus-based system is built to integrate bibliography data from different sources. The experimental results show that our method improves the recommendation performances over other methods.     

Metropolitan patenting, inventor agglomeration and social networks: A tale of two effects

We investigate the separate effects on metropolitan patenting of inventor agglomeration and the structure of social networks linking inventors within and across metropolitan areas in the United States between 1977 and 2002. Using patent data we have been able to assign a metropolitan location to individual inventors, link inventors who have co-authored patents, and characterize the structural features of the networks linking inventors. Our main question concerns the relative importance of salient features of the social networks linking inventors on metropolitan patenting productivity. We find that agglomerative features of metropolitan areas are more important determinants of metropolitan patenting productivity than structural feature of the inventive networks. While the aggregation of isolated inventors correlates positively with patenting productivity, density of connections correlates negatively.     

力学反问题研究中的人工神经网络方法

随着力学学科的不断发展,力学反问题的神经网络方法研究越来越引起工程人员的重视。力学反问题的神经网络方法在工程实践中也得到了一定的应用。     

A cognitive hierarchy model of learning in networks

This paper proposes a method for estimating a hierarchical model of bounded rationality in games of learning in networks. A cognitive hierarchy comprises a set of cognitive types whose behavior ranges from random to substantively rational. Specifically, each cognitive type in the model corresponds to the number of periods in which economic agents process new information. Using experimental data, we estimate type distributions in a variety of task environments and show how estimated distributions depend on the structural properties of the environments. The estimation results identify significant levels of behavioral heterogeneity in the experimental data and overall confirm comparative static conjectures on type distributions across task environments. Surprisingly, the model replicates the aggregate patterns of the behavior in the data quite well. Finally, we found that the dominant type in the data is closely related to Bayes-rational behavior.     

Exploratory analyses of relief and development operations using social networks

We use social networks to explore how structural factors affect humanitarian organizations’ performance in relief and development operations. Analyses of two recent humanitarian disasters show that having pre-established partnerships among implementers, a central coordinator, high connectivity, and few structural holes facilitates coordination and improves performance. Similarly, cost efficiency analyses of 757 development programs reveal that (i) high performance of donors and beneficiaries is positively related to the amount of implementers they interact with, and (ii) programs connected by common actors are more cost efficient. Finally, short path lengths and frequent connections among actors or programs improve performance in humanitarian operations.     

县域城镇化水平差异的人工神经网络测定——以陕西省为例

在已有研究的基础上,运用人工神经网络的理论和方法,构建了BP神经网络模型,并对2008年陕西省83个县的县域城镇化水平进行了测度。结果表明:陕西省县域城镇化水平存在显著分异,评价结果与专家的判断基本一致。陕北、关中地区县域城镇化水平较高,陕南则相对较低,各自区内差异明显。可见,BP神经网络模型运用于县域城镇化水平简单、实用,且有效避免了人工赋权的主观性,具有良好的应用前景。     

An experiment of social learning with endogenous timing

In this paper, we experimentally investigate a social learning model with endogenous timing. Specifically, we focus on a model, in which two subjects are supposed to make a binary decision. One alternative is a safe action with a fixed payoff, while the other alternative is a risky action. The subjects can make their decisions in three stages. The safe action is reversible, but the risky action is not. A subject who delays his decision can observe the decision of the other subject in the earlier stages, and as a result, acquire more information. We show that players do delay their decisions in order to obtain more information. Furthermore, they delay especially when their private information does not particularly support the risky action. We also find evidence which suggests that risk aversion plays an important role in timing decisions, often leading to ex post inefficient outcomes.     

Coordination in the El Farol Bar problem: The role of social preferences and social networks

In this paper, we continue the pursuit of the self-coordination mechanism as studied in the El Farol Bar problem. However, in addition to efficiency (the optimal use of the public facility), we are also interested in the distribution of the public resources among all agents. Hence, we consider variants of the El Farol Bar problem, to be distinguished from many early studies in which efficiency is the only concern. We ask whether self-coordinating solutions can exist in some variants of the El Farol Bar problem so that public resources can be optimally used with neither idle capacity nor congestion being incurred and, in the meantime, the resources can be well distributed among all agents. We consider this ideal situation an El Farol version of a “good society”. This paper shows the existence of a positive answer to this inquiry, but the variants involve two elements, which were largely left out in the conventional literature on the El Farol Bar problem. They are social networks and social preferences. We first show, through cellular automata, that social networks can contribute to the emergence of a “good society”. We then show that the addition of some inequity-averse agents can even guarantee the emergence of the “good society”.     

Leaders emergence in artificial populations: the role of networks

Social networks can be viewed as graphs in their basic form, but many networks of interest include positive and negative affect relations among nodes. The structural balance theory originally developed by Heider suggests how nodes may locally modify their relationships to maintain a sort of balance within sets of nodes. We analyze a model of leadership emergence in a social network and extend it by introducing structural balance among members when modeling the attitude toward the leader. This approach takes into account some of the mutual relationships among co-workers, including the adaptation process to achieve the balance. This component helps to explore differently the bounded rationality of agents when interacting, and prove the difficulty of finding a rapid and smooth covergent path to a social stable equilibrium.