Seasonal Time-Series Model using Particle Swarm Optimization for Broadband Data Payload Prediction

Time-series prediction in telecommunication industry has been used as the basic process of decision making of broadband data selling in order to increase the company's revenue. The conventional methods such as ARIMA, SARIMA and OLS are good methods to predict time series data, but PSO is a better method than those conventional models. PSO was first introduced by Kennedy and Eberhart on 1995 as a part of swarm intelligence algorithm, inspired by the behaviour of birds ( particle) who interacting at each other in a certain velocity and position by following the movement of its pack. In this paper, PSO is challenged to estimate the parameters in the distributed-lag prediction formula with two time lags using seasonal time series of hourly broadband data payload from 1st to 7th December 2014. Simulation result show that PSO give minimum error of Mean Absolute Error ( MAE) than the conventional models ARIMA, SARIMA, and OLS. The PSO can be used as a recommendation to predict seasonal time series of broadband data payload.