Forecasting Iowa gaming volume: A comparison of four time series forecasting methods

This study investigated the forecasting abilities of four forecasting techniques—(a) autoregressive integrated moving average (ARIMA), (b) artificial neural networks (ANNs), (c) simple moving average (SMA), and (d) the naïve method—as they apply to the Iowa monthly time series of slot coin-in and table drop. Mean squared error and mean absolute percentage error were adopted as evaluation criteria to compare the forecasting abilities of these various techniques. The results indicated that SMA outperformed the other three methods, which extends the conclusions of the M-competition to time series in the gaming field. Meanwhile, the ANN technique introduced without any modification was incapable of replacing the standing of ARIMA in the practice of gaming forecasting. This study is the first attempt to explore the forecasting abilities of four prevailing forecasting models based on gaming practices. It provides researchers and practitioners with a guide to and insights into the application of the ANN technique in gaming forecasting, selection of forecasting method, and effectiveness of the model for different horizons of gaming forecasting