Int. J. Mobile Network Design and Innovation, Vol. 1, No. 1, 2005 Copyright © 2005 Inderscience Enterprises Ltd. 18 A neural network-based approach for predicting connectivity in wireless networks Mahdi Nasereddin,* Abdullah Konak and Michael R. Bartolacci Penn State Berks, PO Box 7009, Reading, PA 19610-6009, USA E-mail: mxn16@psu.edu E-mail: konak@psu.edu E-mail: mrb24@psu.edu *Corresponding author Abstract: This paper proposes a Connectivity Decision Support System based on connectivity maps generated by a neural network approach. The proposed approach creates a coverage map based on the signal strengths from active wireless users. These data are used to train a neural network to predict the signal strengths or coverage for locations for which no active user is reporting. In other words, a neural network fills in gaps in a coverage map for a given network connection point. Keywords: wireless networks; connectivity; neural networks; artificial intelligence. Reference to this paper should be made as follows: Nasereddin, M., Konak, A. and Bartolacci, M.R. (2005) ‘A neural network-based approach for predicting connectivity in wireless networks’, Int. J. Mobile Network Design and Innovation, Vol. 1, No. 1, pp.18–23. Biographical notes: Mahdi Nasereddin is an Assistant Professor of Information Sciences and Technology at the Pennsylvania State University-Berks Campus. He received his BS, MS and PhD in Industrial Engineering from the University of Central Florida. His current research interest is in the application of artificial intelligence, simulation metamodelling, simulation optimisation and experimental design. He is a member of INFORMS. Abdullah Konak is an Assistant Professor of Information Sciences and Technology at Penn State Berks. He received his BS degree in Industrial Engineering from Yildiz Technical University, Istanbul, Turkey, MS in Industrial Engineering from Bradley University and PhD in Industrial Engineering from the University of Pittsburgh. Previous to this position, he was an Instructor in the Department of Systems and Industrial Engineering at Auburn University for two years. His current research interest is in the application of operations research techniques to complex problems, including topics such as telecommunications network design, network reliability analysis/optimisation, facilities design and data mining. He is a member of IIE and INFORMS. Michael R. Bartolacci is an Associate Professor of Information Sciences and Technology at Penn State Berks. He received an undergraduate degree in Engineering from Lafayette College, an MBA and a PhD in Industrial Engineering from Lehigh University. His current research interests include telecommunications modelling and analysis, cultural aspects of telecommunication systems and policies and customer relationship management systems. He is a member of INFORMS and AIS. 1 Introduction The maintenance of connectivity between a wireless user and its communicating network is a key issue in both the design of network and its day-to-day operations. Connectivity is loosely defined as the ability to connect to a network connection point with some minimum threshold of signal strength and quality that allows ongoing communication. Connectivity for a wireless network has been traditionally equated with and marketed as the term ‘coverage’. The authors propose a neural network-based approach for rapidly predicting connectivity across an entire region that takes changing patterns of user density and locations, propagation conditions and similar factors into account. The modelling or prediction of connectivity in network design for a fixed infrastructure wireless network, such as cellular phone network, usually involves the gathering of signal strength and quality information for a set of test points. Such information is then used for network connection point placement to ensure adequate coverage/connectivity for users in a given region of service. The number and distribution of such test points depend upon the size of the service region as well as its