Prediction of Network Delay with Variable Standard Deviation, Skewness and Kurtosis

This paper is focused on prediction of the network delay using wedged moving average (WMA) and least linear squares (LLS). The main task is estimation of predictions accuracy at variable moments and autocorrelation of network delay. Samples of network delay with defined moments and exponential autocorrelation are generated by program generator. Three fields of network delay samples are obtained: with variable standard deviation and autocorrelation, with variable skewness and autocorrelation, with variable kurtosis and autocorrelation. Numerical simulation of network delay prediction is performed. The accuracy of prediction is estimated by mean absolute error (MAE). Parameter of predictor with less prediction error is fixed. Simulation results illustrate the correlation of WMA and LLS predictors accuracy with moments of network delay frequency distribution and autocorrelation