International Journal of Modern Physics C

Economy correlations between the 19 richest countries are investigated through their Gross Domestic Product increments. A distance is defined between increment correlation matrix elements and their evolution studied as a function of time and time window size. Unidirectional and Bidirectional Minimal Length Paths are generated and analyzed for different time windows. A sort of critical correlation time window is found indicating a transition for best observations. The mean length path decreases with time, indicating stronger correlations. A new method for estimating a realistic minimal time window to observe correlations and deduce macroeconomy conclusions from such features is thus suggested.

Using data from a sample of 28 representatives countries, we propose a classification of currency crises consequences based on the ultrametric analysis of the real exchange rate movements time series, without any further assumption. By using the matrix of synchronous linear correlation coefficients and the appropriate metric distance between pairs of countries, we were able to construct a hierarchical tree of countries. This economic taxonomy provides relevant information regarding liaisons between countries and a meaningful insight about the contagion phenomenon.

The aim of this paper is to apply the methods of Symbolic Time Series Analysis (STSA) to a series of inflation from a group of Latin-American economies. Starting with a partition of two inflation regimes, we use data symbolization for identifying temporal patterns. Afterwards the statistical information obtained from the patterns is used to estimate the parameters of a nonlinear model proposed by Brida (2000).¹ We compare the performance of the model against a naive benchmark predictor to verify its power to anticipate the qualitative behavior of the inflation time series. When the use of STSA is made through pure optimization criteria, the performance of the model is poor. However, when the partition of the space of states is made according to economics intuition, the performance of the model increases considerably.

We present a multi-agent system where typical labor market mechanisms emerge. Based on a few simple rules, our model allows for different interpretative paradigms to be represented and for different scenarios to be tried out. We thoroughly explore the space of possible strategies both for those unemployed and for companies and analyze the trade-off between these strategies regarding global social and economical indicators.

We developed a mean-field approach to the Minority Game, that allows us to reproduce the behavior of the model in the phase in which a so-called "dynamics of period two" is present. This dynamics describes the situation where the model is controlled by the presence of crowds of agents participating in the game. Our approach is based on the hypothesis that we can introduce states representative of the system, in such a form that averages over time can be replaced by averages over those states. The main idea is to work with virtual agents, rather than working with the actual set of agents of a particular game. Virtual agents are built from all the possible pairs of the strategies available in the model (the Full Strategy Space FSS). Moreover, we define an ensemble of microstates and, thereafter, states compatible with the specifications of the game. In this work we explain in detail how to introduce these elements, and how to actually calculate the ensemble of states and microstates. We have developed one generalization of the Minority Game as an attempt to make the model more realistic, by introducing interactions among the agents. We also discuss and explain the adequate ensemble of states and microstates for that generalization.

We analyze a Lotka-Volterra-like social lattice of agents interacting among themselves in collaborative or competitive scenarios. Dynamic characteristics of the model are derived, equilibrium points of the system are found. For the nearest neighbor interaction structure case, which depends on a γ₁ parameter, also bifurcations are explored allowing us to understand how different levels of collaboration or competition leads the system to different configurations. We found that, within the range -α/4 < γ₁ < α/4 the system tend to an uniform configuration, and when competition becomes aggressive, γ₁ > α/4, agents tend to polarized opinions and contrarians agents appear naturally. Higher neighbor structures have been also simulated, for which contrarian patterns also emerge dynamically. In all cases the initial conditions of the lattice are randomly taken. Hung and conflictive scenarios are discussed.

A model for the displacement of self-driven organisms is studied by means of extensive computer simulations. Local interactions influenced by noisy communications among organisms, leads to the onset of collective motion at low noise levels. When the noise is increased the system undergoes first-order transitions into disordered states of motion. We have also studied the relaxation process between these states. By fitting the time dependence of the order parameter when the system is annealed from a state below coexistence to another above it, we conclude that the relaxation can be well described by means of a stretched exponential. In this way the characteristic relaxation times are obtained.

We reinvestigate the model of Bonabeau et al.¹ of self-organizing social hierarchies by including a distribution of attractive sites. Agents move randomly except in the case where an attractive site is located in its neighborhood. We find that the transition between an egalitarian society at low population density and a hierarchical one at high population density strongly depends on the distribution and percolation of the valuable sites. We also show how agent diffusivity is closely related to social hierarchy.

A model for opinion formation in a social group, based on the Theory of Social Impact developed by Latané, is studied by means of numerical simulations. Interactions among the members of the group, as well as with a strong leader competing with the mass media, are considered. The model exhibits first-order transitions between two different states of opinion, which are supported by the leader and the mass media, respectively. The social inertia of the group becomes evident when the opinion of the leader changes periodically. In this case two dynamic states are identified: for long periods of time, the group follows the changes of the leader but, decreasing the period, the opinion of the group remains unchanged. This scenery is suitable for the ocurrence of dynamic phase transitions.

In the present work we studied the room evacuation problem using the social force model introduced by Helbing and coworkers. The "faster is slower" effect induced by panic was analyzed. It could be explained in terms of increasing mean clogging delays which shows a strong correlation with certain structures that we call "blocking clusters". Also a steady state version of the problem was implemented. It shows that, from a macroscopic point of view, the optimal evacuation efficiency correspond to the state at which the difference between the system desire force minus the system granular force is maximum.

We studied the dynamical behavior of a mobile communication network through a traditional Ising model. In the mobile communication network, each cell has N channels to process the calls, and they arrive with a given rate, having different lengths. In our Ising model we introduce cooperation among cells. This mechanism is considered in the normal commercial operation today. Through our simulation we evaluate the (virtual) temperature, magnetization, and the critical parameters of the network. The description of a mobile communication network, using a well-known theoretical model, is an important tool to improve the performance of these networks.

A Bayesian network is a directed acyclic graph in which each node represents a variable and each arc a probabilistic dependency; they are used to provide: a compact form to represent the knowledge and flexible methods of reasoning. Obtaining it from data is a learning process that is divided in two steps: structural learning and parametric learning. In this paper we define an automatic learning method that optimizes the Bayesian networks applied to classification, using a hybrid method of learning that combines the advantages of the induction techniques of the decision trees (TDIDT-C4.5) with those of the Bayesian networks. The resulting method is applied to prediction in health domain.