Soft computing

About: Soft computing is a research topic. Over the lifetime, 6710 publications have been published within this topic receiving 118508 citations.

A neural network and fuzzy rule base hybrid for pattern classification

Abstract: This paper presents a novel hybrid of the two complimentary technologies of soft computing viz. neural networks and fuzzy logic to design a fuzzy rule based pattern classifier for problems with higher dimensional feature spaces. The neural network component of the hybrid, which acts as a pre-processor, is designed to take care of the all-important issue of feature selection. To circumvent the disadvantages of the popular back propagation algorithm to train the neural network, a meta-heuristic viz. threshold accepting (TA) has been used instead. Then, a fuzzy rule based classifier takes over the classification task with a reduced feature set. A combinatorial optimisation problem is formulated to minimise the number of rules in the classifier while guaranteeing high classification power. A modified threshold accepting algorithm proposed elsewhere by the authors (Ravi V, Zimmermann H.-J. (2000) Eur J Oper Res 123: 16–28) has been employed to solve this optimization problem. The proposed methodology has been demonstrated for (1) the wine classification problem having 13 features and (2) the Wisconsin breast cancer determination problem having 9 features. On the basis of these examples the results seem to be very interesting, as there is no reduction in the classification power in either of the problems, despite the fact that some of the original features have been completely eliminated from the study. On the contrary, the chosen features in both the problems yielded 100% classification power in some cases.

A soft-computing based rough sets classifier for classifying IPO returns in the financial markets

Abstract: In the financial markets, due to limitations of the noise caused continuously by changing market conditions and environments, and a subjective sentiment or other factors unrelated to expected returns on investment decision-making of investors, there is a growing consensus designing and employing a variety of soft computing systems to remedy the aforementioned existing problems objectively and intelligently Previously, many researchers have long used statistical methods for handling the related problems of investment markets However, these conventional methods become more complex when relationships in the input/output dataset are nonlinear Nevertheless, statistical techniques always rely on the assumptions on linear separability, multivariate normality, and independence of the predictive variables; unfortunately, many of the common models of treating the financial markets problems violate these assumptions Therefore, to reconcile the existing shortcomings, this study offers three hybrid models based on a rough sets classifier to extract decision rules and aid making investment decision for the market investors The proposed hybrid models include three differently integrated models for solving IPO (Initial Public Offerings) returns problems of the financial markets: (1) Experiential Knowledge (EK)+Feature Selection Method (FSM)+Minimize Entropy Principle Approach (MEPA)+Rough Set Theory (RST)+Rule Filter (RF), (2) EK+Decision Trees (DT)-C45+RST+RF, and (3) EK+FSM+RST+RF The proposed hybrid models are illustrated by examining an IPO dataset for publicly traded firms The experimental results indicate that the proposed hybrid models outperform the listing methods in accuracy, number of attributes, standard deviation, and number of rules Furthermore, the proposed hybrid models generate comprehensible rules readily applied in knowledge-based systems for investors Meaningfully, the study findings and implications are of value to both academicians and practitioners

Recent achievements in sensor localization algorithms

Abstract: Internet Of Things (IOT) is an inevitable result of the evolution of communication and manufacturing of small low power and effective micro electro mechanical systems (MEMS). Wireless Sensor Network (WSN) is Self organized collected sensors that communicate to each other randomly through waves. Node localization is a major challenge for most of WSN applications due to the difficult environments through which nodes are distributed such as underwater, fires, volcanoes, animal habitat or Battlefields. Moreover, using Global Positioning System (GPS) is costly high and not feasible in indoors. Range based algorithms such as Angle Of Arrival, Time Of Arrival and Received Signal Strength Indicator are alternative for GPS with acceptable accuracy, however they require additional hardware. Range free algorithms such as centroid, approximate point-in-triangulation and distance vector-hop algorithms are economic but less accurate. Lack of accuracy is still an issue in traditional localization algorithms which guide researchers recently to propose optimization techniques with the ability to improve the accuracy of localization based on soft computing algorithms like genetic algorithms, fuzzy logic, neural network and deep learning. The objective of this paper is to give an over view of WSN applications and challenges highlighting the localization problem. The paper also, proposes taxonomy that classifies variant localization algorithms. Moreover, presents a study of different traditional algorithms as well as their improvements based on soft computing techniques. Moreover, the paper summarizes the challenges of localization and represents research points for future work.

Traffic speed prediction method for urban networks — an ANN approach

Abstract: The paper proposes a traffic speed prediction algorithm for urban road traffic networks. The motivation of the prediction is to provide short time forecast in order to support ITS (Intelligent Transport System) functionalities, such as traveler information systems, route guidance (navigation) systems, as well as adaptive traffic control systems. A potential and efficient solution to this problem is the application of a soft computing method. Namely, an artificial neural network (ANN) is used for the forecast by involving the measured speed patterns. The ANN is trained by using data produced by Vissim (a microscopic road traffic simulator) simulations. The proposed algorithm is developed and analyzed on a real-word test network (part of downtown in Budapest).

Enhanced WiFi localization system based on Soft Computing techniques to deal with small-scale variations in wireless sensors

Abstract: The framework of this paper is robot localization inside buildings by means of wireless localization systems. Such kind of systems make use of the Wireless Fidelity (WiFi) signal strength sensors which are becoming more and more useful in the localization stage of several robotic platforms. Robot localization is usually made up of two phases: training and estimation stages. In the former, WiFi signal strength of all visible Access Points (APs) are collected and stored in a database or WiFi map. In the latter, the signal strengths received from all APs at a certain position are compared with the WiFi map to estimate the robot location. Hence, WiFi localization systems exploit the well-known path loss propagation model due to large-scale variations of WiFi signal to determine how closer the robot is to a certain AP. Unfortunately, there is another kind of signal variations called small-scale variations that have to be considered. They appear when robots move under the wavelength @l. In consequence, a chaotic noise is added to the signal strength measure yielding a lot of uncertainty that should be handled by the localization model. While lateral and orientation errors in the robot positioning stage are well studied and they remain under control thanks to the use of robust low-level controllers, more studies are needed when dealing with small-scale variations. Moreover, if the robot can not use a robust low-level controller because, for example, the environment is not organized in perpendicular corridors, then lateral and orientation errors can be significantly increased yielding a bad global localization and navigation performance. The main goal of this work is to strengthen the localization stage of our previous WiFi Partially Observable Markov Decision Process (POMDP) Navigation System with the aim of dealing effectively with small-scale variations. In addition, looking for the applicability of our system to a wider variety of environments, we relax the necessity of having a robust low-level controller. To do that, this paper proposes the use of a Soft Computing based system to tackle with the uncertainty related to both the small-scale variations and the lack of a robust low-level controller. The proposed system is actually implemented in the form of a Fuzzy Rule-based System and it has been evaluated in two real test-beds and robotic platforms. Experimental results show how our system is easily adaptable to new environments where classical localization techniques can not be applied since the AP physical location is unknown.