Reinforcement and Non-Reinforcement Machine Learning Classifiers for User Movement Prediction

Mobile context-aware applications are required to sense and react to changing environment conditions. Such applications, usually, need to recognize, classify, and predict context in order to act efficiently, beforehand, for the benefit of the user. In this chapter, the authors propose a mobility prediction model, which deals with context representation and location prediction of moving users. Machine Learning (ML) techniques are used for trajectory classification. Spatial and temporal on-line clustering is adopted. They rely on Adaptive Resonance Theory (ART) for location prediction. Location prediction is treated as a context classification problem. The authors introduce a novel classifier that applies a Hausdorff-like distance over the extracted trajectories handling location prediction. Two learning methods (non-reinforcement and reinforcement learning) are presented and evaluated. They compare ART with Self-Organizing Maps (SOM), Offline kMeans, and Online kMeans algorithms. Their findings are very promising for the use of the proposed model in mobile context aware applications.

The Applications of Automata in Game Theory

In game theory, presenting players with strategies directly affects the performance of the players. Utilizing the power of automata is one way for presenting players with strategies. In this chapter, the authors studied different types of automata and their applications in game theory. They found that finite automata, adaptive automata, and cellular automata are widely adopted in game theory. The applications of finite automata are found to be limited to present simple strategies. In contrast, adaptive automata and cellular automata are intensively applied in complex environment, where the number of interacted players (human, computer applications, etc.) is high, and therefore, complex strategies are needed.

Evolutionary Control of Helicopter Hovering Based on Genetic Programming

Computational intelligence techniques such as neural networks, fuzzy logic, and hybrid neuroevolutionary and neuro-fuzzy methods have been successfully applied to complex control problems in the last two decades. Genetic programming, a field under the umbrella of evolutionary computation, has not been applied to a sufficiently large number of challenging and difficult control problems, in order to check its viability as a general methodology to such problems. Helicopter hovering control is considered a challenging control problem in the literature and has been included in the set of benchmarks of recent reinforcement learning competitions for deriving new intelligent controllers. This chapter shows how genetic programming can be applied for the derivation of controllers in this nonlinear, high dimensional, complex control system. The evolved controllers are compared with a neuroevolutionary approach that won the first position in the 2008 helicopter hovering reinforcement learning competition. The two approaches perform similarly (and in some cases GP performs better than the winner of the competition), even in the case where unknown wind is added to the dynamic system and control is based on structures evolved previously, that is, the evolved controllers have good generalization capability.

Intention Prediction Mechanism in an Intentional Pervasive Information System

The development of pervasive technologies has allowed the improvement of services availability. These services, offered by Information Systems (IS), are becoming more pervasive, i.e., accessed anytime, anywhere. However, those Pervasive Information Systems (PIS) remain too complex for the user, who just wants a service satisfying his needs. This complexity requires considerable efforts from the user in order to select the most appropriate service. Thus, an important challenge in PIS is to reduce user’s understanding effort. In this chapter, the authors propose to enhance PIS transparency and productivity through a user-centred vision based on an intentional approach. They propose an intention prediction approach. This approach allows anticipating user’s future requirements, offering the most suitable service in a transparent and discrete way. This intention prediction approach is guided by the user’s context. It is based on the analysis of the user’s previous situations in order to learn user’s behaviour in a dynamic environment.

Context-Based Grouping and Recommendation in MANETs

The authors propose in this chapter a context grouping mechanism for context distribution over MANETs. Context distribution is becoming a key aspect for successful context-aware applications in mobile and ubiquitous computing environments. Such applications need, for adaptation purposes, context information that is acquired by multiple context sensors distributed over the environment. Nevertheless, applications are not interested in all available context information. Context distribution mechanisms have to cope with the dynamicity that characterizes MANETs and also prevent context information from being delivered to nodes (and applications) that are not interested in it. The authors’ grouping mechanism organizes the distribution of context information in groups whose definition is context based: each context group is defined based on a criteria set (e.g. the shared location and interest) and has a dissemination set, which controls the information that can be shared in the group. They propose a personalized and dynamic way of defining and joining groups by providing a lattice-based classification and recommendation mechanism that analyzes the interrelations between groups and users, and recommend new groups to users, based on the interests and preferences of the user.

On the Use of Optimal Stopping Theory for Secret Sharing Scheme Update

The location privacy issue has been addressed thoroughly so far. Cryptographic techniques, k-anonymity-based approaches, spatial obfuscation methods, mix-zones, pseudonyms, and dummy location signals have been proposed to enhance location privacy. In this chapter, the authors propose an approach, called STS (Share The Secret) that segments and distributes the location information to various, non-trusted, entities from where it will be reachable by authenticated location services. This secret sharing approach prevents location information disclosure even in situation where there is a direct observation of the target. The proposed approach facilitates end-users or location-based services to classify flexible privacy levels for different contexts of operation. The authors provide the optimal thresholds to alter the privacy policy levels when there is a need for relaxing or strengthening the required privacy. Additionally, they discuss the robustness of the proposed approach against various adversary models. Finally, the authors evaluate the approach in terms of computational and energy efficiency, using real mobile applications and location update scenarios over a cloud infrastructure, which is used to support storage and computational tasks.

Enhancing Location Privacy in WSN

Wireless Sensor Networks (WSNs) receive significant attention due to the wide area of applications: environment monitoring, tracking, target detection, etc. At the same time, in some cases, the captured information from the WSN might be considered as private, for example, location of an important asset. Thus, security mechanisms might be essential to ensure the confidentiality of the location of the information source. In this chapter, the authors present an approach called iHIDE (information HIding in Distributing Environments) to enable source-location privacy in WSNs. iHIDE adopts a non-geographical, overlay routing method for packet delivery. This chapter presents the architecture and assesses its performance through simulation experiments, providing comparisons with relative approaches.

A Forest Fire Detection System

Forest fires cause immeasurable damages to indispensable resources for human survival, destroy the balance of earth ecology, and worst of all they frequently cost human lives. In recent years, early fire detection systems have emerged to provide monitoring and prevention of the disasterous forest fires. Among them, the Meleager1 system aims to offer one of the most advanced and integrated technology solutions for fire protection worldwide by integrating several innovative features. This chapter outlines one of the major components of the Meleager system, that is the visual fire detection sybsystem. Groundbased visible range PTZ cameras monitor the area of interest, and a low level decision fusion scheme is used to combine individual decisions of numerous fire detection algorithms. Personalized alerts and induced feedback is used to adapt the detection process and improve the overall system performance.

Methodology for Epilepsy and Epileptic Seizure Recognition using Chaos Analysis of Brain Signals

A novel combination of chaotic features and Adaptive Neuro-Fuzzy Inference System (ANFIS) is proposed for epileptic seizure recognition. The non-linear dynamics of the original EEGs are quantified in the form of the Hurst exponent (H), Correlation dimension (D2), Petrosian Fractal Dimension (PFD), and the Largest lyapunov exponent (?). The process of EEG analysis consists of two phases, namely the qualitative and quantitative analysis. The classification ability of the H, D2, PFD, and ? measures is tested using ANFIS classifier. This method is evaluated with using a benchmark EEG dataset, and qualitative and quantitative results are presented. The inter-ictal EEG-based diagnostic approach achieves 98.6% accuracy with using 4-fold cross validation. Diagnosis based on ictal data is also tested in ANFIS classifier, reaching 98.1% accuracy. Therefore, the method can be successfully applied to both inter-ictal and ictal data.

Learning Methodologies to Support E-Business in the Automated Negotiation Process

The automated negotiation performed by a software agent is investigated in order to improve the benefits compared to a humane face-to-face negotiation. The profitability of e-business applications can be increased by the support of automated negotiation tools. This research proposes a set of learning methodologies to support both the suppliers’ and customers’ negotiation activities. The learning methodologies are based on Q-learning technique, which is able to evaluate the utility of the actions without a model of the environment. The context regards one-to-many negotiation and multi-issues (volume, price, and due date). A simulation environment is developed to test the proposed methodologies and evaluate the benefits compared to a negotiation approach without learning support tool. The simulations are conducted in several market conditions, and a proper statistical analysis is performed. The simulation results show that the proposed methodologies lead to benefits both for suppliers and customers when both the opponents adopt the learning approach.