An Alternative Algorithm for ARIMA Model Selection

Modelling and forecasting of time series data are essential components of data science and big data analytics. Over the years, the Box-Jenkins Methodology has gained tremendous popularity and use for modelling univariate time series. However, the plots of the Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) often give conflicting and biased pictures, which makes it di cult for proper Autoregressive Integrated Moving Average (ARIMA) model identification or selection. In this paper, we propose an alternative algorithm based on the principles of Cartesian products of sets in mathematics. The practicality of this algorithm was demonstrated with application to univariate ARIMA modelling of Nigerian stock price data. The results show that this algorithm is flexible, modular and widely applicable for quick ARIMA model selection.