Prediction of demand/usage patterns for services in telecommunications using fuzzy neural networks

This paper presents a generalized algorithm for forecasting the demand for various services in a telecommunications network with a degree of confidence. Knowledge of demand and traffic patterns between geographic areas is essential for network design and planning. Neural networks have exhibited the properties of online adaptation and interpolating within and outside the scope of the input data set. This ability to learn new samples of data while retaining previously acquired knowledge about the data set, makes neural networks an excellent tool for time-series analysis. In this work, a neural network with fuzzy learning rules is employed for time-series prediction, and the results obtained are compared with those generated using linear regression. These two solutions are fed as inputs to a fuzzy controller that determines the degree of certainty with which the integrated forecast can be made, based on the correlation between the two individually obtained solutions