Comparison of Forecasting ability of Sama Circular Model, ARIMA and Decomposition Techniques

Abstract

<p>Time series data comprises several components; Trend, Seasonal variations, cyclical variations and irregular variations. These series follow irregular wave-like patterns. This type of data is common in the fields of, Meteorology, Agriculture, Share Market Economic, Education, Healthcare and more.&nbsp; The Decomposition technique and the Auto Regressive Integrated Moving Average (ARIMA)/Seasonal Auto Regressive Integrated Moving Average (SARIMA) are the widely applied methods for forecasting such a series. Yet these techniques are unable to model the cyclical variation and they have some other weaknesses. According to the literature, modelling cyclical variation is highly important and crucial. Some researchers have attempted the Artificial Neural Network for the purpose, yet the success of them was doubtful. There were no Statistical techniques for the purpose. The Sama Circular Model (SCM) is a recently joined member to the family of forecasting techniques, developed on Newton&rsquo;s law of Circular Motion, Fourier transformation and Least Square Estimation. Indeed it is a frequency domain model. The SCM is capable in capturing all the components of a time series; Trend, Seasonal and Cyclical. Hence it was intended to compare the forecasting ability of the SCM with Decomposition techniques and ARIMA/SARIMA. Monthly female unemployment rates of Australia for forty years from 1978 were used for model testing. The goodness of fit tests and measurements of errors were used in model validation and verification. The SCM was superior to the ARIMA/SARIMA and Decomposition techniques.</p>