International Journal of Advanced Networking Applications (IJANA) ISSN No. : 0975-0290 14 A Literature Review on Recent Advances in Neuro-Fuzzy Applications Nidhi Arora U.V. Patel College of Engineering, Ganpat University, Kherva, Gujarat, India Email: aroranidhig@rediffmail.com Jatinderkumar R. Saini Narmada College of Computer Application, Bharuch, Gujarat, India Email: saini_expert@yahoo.com ----------------------------------------------------------------------ABSTRACT-------------------------------------------------------------- Traditional algorithmic approaches are not suitable for solving today’s business problems. Neuro-Fuzzy systems have recently become popular and promising choice among researchers in attempt to solve complex problems faced in business. The paper presents a brief review of most recent applications in business aimed at knowing future events in advance specifically employing neuro-fuzzy approach. The neuro-fuzzy systems designed and developed during 2011 and 2014 have been studied with the intention to explore the recent developments. Prominent applications in some wide-spread domains are considered with an outlook of their capabilities in respective domains. Keywords - Fuzzy logic, Hybrid systems, Neural networks, Traditional methods, Uncertainty I. INTRODUCTION In recent years, numerous studies have been carried out which revealed the limitations of traditional methods in dealing with problems of today. The methods currently used in business are somewhere incapable of prediction of future events to act proactively due to vagueness and mass of the data or due to complexity of the problem. Everything cannot be represented precisely due to unavailability of proper information, lack of information or unclear information. It is evident that traditional computing methods are inefficient in such situations. A significant research contribution in finding solution to these problems has made companies to see a gradual shift from traditional methods to advanced systems. The reason behind the motivation is obvious; the common characteristics of business problems like non-linear behavior, high degree of uncertainty, lack of precise knowledge etc. A variety of machine learning techniques are being developed for possible application into various fields to experience intelligent information systems. The advancement of technology, success of upcoming methods and their acceptance has given way to modified methods in almost all areas. An emerging class of intelligent machines taking place of traditional methods consists of fuzzy logic and neural networks. Fuzzy logic provides a mathematical foundation for dealing with situations full of uncertainties by simulating human perception for understanding linguistic attributes while neural network mimics human beings in the process of learning, thinking and adaptation. A combination of these two techniques called neuro-fuzzy systems has the potential to get the advantages of both leaving behind their limitations. In a hybrid model, neural network learning algorithms are fused with fuzzy reasoning of fuzzy logic. Each of the components plays their own role; neural network determine the values of parameters while if-then rules are handled by fuzzy logic. A fuzzy inference system uses human expertise by storing required knowledge in its rule-base, and then performs fuzzy reasoning on the input to infer the overall output value. Fuzzy logic and neural networks complement each other in developing intelligent systems. Neural networks form low-level computational structures and can deal with raw data while, fuzzy logic sits on a higher level and uses linguistic information. Looking from the other viewpoint, fuzzy systems lack the ability to learn and cannot adjust to a new environment, but neural networks can learn and generalize. A general problem with neural networks is that they are black box with number of hidden layers whose operation is opaque to user. Integrated together, the resultant neuro-fuzzy system can perform parallel computation and learn like neural networks and can represent human-like knowledge with explanation abilities of fuzzy systems making the overall system more transparent. Researchers from varied domains have attempted and relied on the use of hybrid neural network and fuzzy logic approach. A neuro-fuzzy system has input and output layers, and hidden layers representing membership functions and fuzzy rules, which learns to solve problems full of uncertainty. It is not possible to include all the applications on neuro-fuzzy hybridization, so we restrict the discussion to domains like education and banking & finance where the day to day decisions are affected by customer behavior. II. Existing Applications There are numerous interesting applications that aim to solve complexity of business problems. The advent of technology has opened doors for Artificial Neural Networks (ANN) which have tendency to simplify the programming