Skip to main content
main-content
Top

About this book

Militarized Conflict Modeling Using Computational Intelligence examines the application of computational intelligence methods to model conflict. Traditionally, conflict has been modeled using game theory. The inherent limitation of game theory when dealing with more than three players in a game is the main motivation for the application of computational intelligence in modeling conflict.

Militarized interstate disputes (MIDs) are defined as a set of interactions between, or among, states that can result in the display, threat or actual use of military force in an explicit way. These interactions can result in either peace or conflict. This book models the relationship between key variables and the risk of conflict between two countries. The variables include Allies which measures the presence or absence of military alliance, Contiguity which measures whether the countries share a common boundary or not and Major Power which measures whether either or both states are a major power.

Militarized Conflict Modeling Using Computational Intelligence implements various multi-layer perception neural networks, Bayesian networks, support vector machines, neuro-fuzzy models, rough sets models, neuro-rough sets models and optimized rough sets models to create models that estimate the risk of conflict given the variables. Secondly, these models are used to study the sensitivity of each variable to conflict. Furthermore, a framework on how these models can be used to control the possibility of peace is proposed. Finally, new and emerging topics on modelling conflict are identified and further work is proposed.

Table of Contents

Frontmatter

Chapter 1. Modeling Conflicts Between States: New Developments for an Old Problem

Abstract
This chapter reviews the evolution of academic research on interstate conflicts. In particular, the emphasis is on the methodological development that has happened through the analysis of interstate conflicts as a scientific phenomenon in the political science discipline. Key empirical findings and theoretical contributions that this change has produced are also emphasized. In addition, the chapter makes the case for the use of computational intelligence for modeling conflicts as a way to overcome some limitations of previous work and unpacks the complexity that is inherent in dispute behavior. The data set used for the analysis in this book is also explained in detail.
Tshilidzi Marwala, Monica Lagazio

Chapter 2. Automatic Relevance Determination for Identifying Interstate Conflict

Abstract
This chapter introduces the Bayesian and the evidence frameworks to construct an automatic relevance determination method. These techniques are described in detail, relevant literature reviews were conducted and their use is justified. The automatic relevance determination technique was then applied to determine the relevance of interstate variables that are essential for modeling interstate conflict. Conclusions are drawn and explained within the context of political science.
Tshilidzi Marwala, Monica Lagazio

Chapter 3. Multi-layer Perceptron and Radial Basis Function for Modeling Interstate Conflict

Abstract
This chapter introduces and then compares the multi-layer perceptron neural network to the radial basis function neural network to help understand and predict interstate conflict. These two techniques are described in detail and justified with a review of relevant literature and they are implemented to interstate conflict. The results obtained from the implementation of these techniques demonstrate that the multi-layer perceptron neural network is better at predicting interstate conflict than the radial basis function network. This is mainly due to the cross-coupled chartacteristics of the multi-layer perceptron’s network compared to the radial basis function network.
Tshilidzi Marwala, Monica Lagazio

Chapter 4. Bayesian Approaches to Modeling Interstate Conflict

Abstract
Two Bayesian techniques are described in this chapter and compared for interstate conflict prediction. The first one is the Bayesian technique that applies the Gaussian approximation approach to approximate the posterior probability for neural network weights, given the observed data and the evidence framework to train a multi-layer perceptron neural network. The second one treats the posterior probability as is, and then applies the hybrid Monte Carlo technique to train the multi-layer perceptron neural network. When these techniques are applied to model militarized interstate disputes, it is observed that training the neural network with the posterior probability as is, and applying the hybrid Monte Carlo technique gives better results than approximating the posterior probability with a Gaussian approximation method and then applying the evidence framework to train the neural network.
Tshilidzi Marwala, Monica Lagazio

Chapter 5. Support Vector Machines for Modeling Interstate Conflict

Abstract
Militarized conflict is one of the risks that have a significant impact on society. Militarized interstate dispute is defined as an outcome of interstate interactions, which result either in peace or conflict. The effective prediction of the possibility of conflict between states is an important decision support tool for policy makers. In previous chapters, neural networks were implemented to predict militarized interstate disputes. Support vector machines have proved to be excellent predictors and hence are introduced in this chapter for the prediction of militarized interstate disputes and then compared with the hybrid Monte Carlo trained multi-layer perceptron neural networks. The results demonstrated that support vector machines predict militarized interstate dispute better than neural networks, while neural networks give a more consistent and easy to interpret sensitivity analysis than do support vector machines.
Tshilidzi Marwala, Monica Lagazio

Chapter 6. Fuzzy Sets for Modeling Interstate Conflict

Abstract
This chapter investigates the level of transparency of the Takagi-Sugeno neuro-fuzzy model and the support vector machines model by applying them to conflict management, an application which is concerned with causal interpretations of results. The data set used in this investigation is the militarized interstate disputes dataset obtained from the Correlates of War (COW) project. In this chapter, a support vector machine model is trained to predict conflict. Knowledge from the Takagi-Sugeno neuro-fuzzy model is extracted by interpreting the model’s fuzzy rules and their outcomes. It is found that the Takagi-Sugeno neuro-fuzzy model offers some transparency which helps in understanding conflict management. The Takagi-Sugeno neuro-fuzzy model was compared to the support vector machine model and it was found that even though the support vector machine shows marginal advantage over the Takagi-Sugeno neuro-fuzzy model in terms of predictive capacity, the Takagi-Sugeno neuro-fuzzy model allows for linguistics interpretation.
Tshilidzi Marwala, Monica Lagazio

Chapter 7. Rough Sets for Modeling Interstate Conflict

Abstract
This chapter applies the rough set technique to model militarized interstate disputes. One aspect of modeling using rough sets is the issue of granulizing the input data. In this chapter, two granulization techniques are introduced, implemented, and compared. These are the equal-width-bin and equal-frequency-bin partitioning techniques. The rough set model is also compared to the neuro-fuzzy model introduced in Chap.​ 6. The results obtained demonstrate that equal-width-bin partitioning gives better accuracy than equal-frequency-bin partitioning. However, both techniques were found to give less accurate results than neuro-fuzzy sets. Also, they were found to be more transparent than neuro-rough sets. Furthermore, it is observed that the rules generated from the rough sets are linguistic and easy-to-interpret in comparison with the ones generated from the neuro-fuzzy model.
Tshilidzi Marwala, Monica Lagazio

Chapter 8. Particle Swarm Optimization and Hill-Climbing Optimized Rough Sets for Modeling Interstate Conflict

Abstract
This chapter presents methods to optimally granulize rough set partition sizes using particle swarm optimization and hill climbing techniques. These two methods are then compared to the equal-width-bin partitioning technique. The results obtained demonstrated that hill climbing provides higher forecasting accuracy, followed by the particle swarm optimization method, which was better than the equal-width-bin technique.
Tshilidzi Marwala, Monica Lagazio

Chapter 9. Simulated Annealing Optimized Rough Sets for Modeling Interstate Conflict

Abstract
In this chapter, methods to optimally granulize rough set partition sizes using simulated annealing technique, are proposed. The proposed procedure is applied to model the militarized interstate dispute data. The suggested technique is then compared to the rough set partition method that is based on particle swarm optimization. The results obtained demonstrate that simulated annealing provides higher forecasting accuracies than particle swarm optimization method.
Tshilidzi Marwala, Monica Lagazio

Chapter 10. Genetic Algorithm with Optimized Rough Sets for Modeling Interstate Conflict

Abstract
This chapter presents methods to optimally granulize rough set partition sizes using a genetic algorithm. The procedure is applied to model the militarized interstate dispute data. The procedure is then compared to the rough set partition method that was based on simulated annealing. The results obtained showed that, for the data being analyzed, a genetic algorithm provides higher forecasting accuracy than does the process of simulated annealing.
Tshilidzi Marwala, Monica Lagazio

Chapter 11. Neuro-Rough Sets for Modeling Interstate Conflict

Abstract
This chapter investigated a neuro-rough model –a combination of a Multi-Layered Perceptron (MLP) neural network with rough set theory– for the modeling of interstate conflict. The model was formulated using a Bayesian framework and trained using a Monte Carlo technique with the Metropolis criterion. The model was then tested on militarized interstate dispute and was found to combine the accuracy of the Bayesian MLP model with the transparency of the rough set model. The technique presented was compared to the genetic algorithm optimized rough sets. The presented Bayesian neuro-rough model performed better than the genetic algorithm optimized rough set model.
Tshilidzi Marwala, Monica Lagazio

Chapter 12. Early Warning and Conflict Prevention Using Computational Techniques

Abstract
This chapter reviews our principal findings and their implications for early warning and conflict prevention. The results of all our analyses are integrated to provide a possible single solution for increasing peace in the international system. In this chapter, a control algorithm is created using computational intelligence. The chapter assesses different general theories and approaches to early warning and conflict prevention as well as the role that computational intelligence could play in enhancing international early warning and conflict prevention. Finally, the chapter presents our diagnosis and prognosis for the future of international relations. Special attention is given to the three pillars of Kantian peace – democracy, economic interdependence and international organizations – and how, on the basis of our analyses, the international community should use these three forces to promote and spread peace.
Tshilidzi Marwala, Monica Lagazio

Chapter 13. Conclusions and Emerging Topics

Abstract
The capability to scientifically comprehend the causes of militarized interstate disputes and then to apply this knowledge to build and spread peace in the international context is unquestionably a vital endeavor. Recent advances in the conflict literature have underlined the importance of handling international conflicts as complex phenomena, exhibiting non-linear and complex interactions amongst the relevant militarized interstate dispute variables.
Tshilidzi Marwala, Monica Lagazio

Backmatter

Additional information

Premium Partner

    Image Credits