Abstract-Reinforcement learning (RL) is a valuable learning method when the systems require a selection of control actions whose consequences emerge over long periods for which inputoutput data are not available. In most combinations of fuzzy systems and RL, the environment is considered to be deterministic. In many problems, however, the consequence of an action may be uncertain or stochastic in nature. In this paper, we propose a novel RL approach to combine the universal-function-approximation capability of fuzzy systems with consideration of probability distributions over possible consequences of an action. The proposed generalized probabilistic fuzzy RL (GPFRL) method is a modified version of the actor-critic (AC) learning architecture. The learning is enhanced by the introduction of a probability measure into the learning structure, where an incremental gradient-descent weightupdating algorithm provides convergence. Our results show that the proposed approach is robust under probabilistic uncertainty while also having an enhanced learning speed and good overall performance.Index Terms-Actor-critic (AC), learning agent, probabilistic fuzzy systems, reinforcement learning (RL), systems control.

Abstract:Although, there are many successful E-commerce organizations, it is argued that E-commerce has not reached its full potential. Trust was often cited as the main reason as many customers are still skeptical about some online vendors. Many trust models have been developed, but most are subjective and did not take into account the vagueness and ambiguity of the domain and the specificity of customers. We have developed a model that attempts to identify the information customers expect to find on a vendors website to increase their trust and hence the likelihood of a transaction to take place. The system is supported by an information extraction system that facilitates the information gathering process. In this paper, we present a method based on fuzzy logic to evaluate trust in E-commerce based on the extracted information. We argue that fuzzy logic is suitable for trust evaluation as it takes into account the uncertainties within E-commerce data and like human relationships, trust is often expressed by linguistics terms rather then numerical values. The results of the system validation using two case studies are also presented.

No abstract

Edgbaston, B15 2TT, UK Salford, A45 4wT, UK S.Nefti-Meziani&alford. ac. uk M.Oussalah@bham.ac.uk Abstract -Artificial Neural Networks (AMVs) are becoming increasingly popular for solving complex problems, as they can behave quite well at solving proLlems that don't have an algorithmic solution or for which the algorithmic solution is too complex to be found. In railway systems, the problem of predicting the sJ1steni malfinctions. or equivalently, railway safety is ofparamount interest for most of railway companies. Traditional ways of predicting railway safe@ are very expensive in terms of time consuming, which make them inefficient under certain circumstances. This paper advocates the use of AMVs architecture to handle the safety problem. By taking irregularities in the positioning of the rails as input to the ANN the ANNcanpredict the safety ratio of the rails.In order to reduce the dimensionality of inputs data a wavelet transformation technique has been employed. Drfferent neural network structures are created and lheir performances both in terms of mean squared error and correlation coeficient have been evaluated to$nd om the best structure for predicting railway safety. The experiments showed that when the model is trained cw a dataset subset and then tested on different subser. it perjormed satisfactorily and con predict the desired output with a very low error factor.

No abstract