# Simple moving average

The second ad-hoc method is Simple moving average, in which previous values are used in order to find the most suitable parameter that gives the lowest forecasting error. The crucial part in this method is correct choice of number of periods taken in the forecast. Weatherford and Kimes (2003) were testing 2 – 8 periods and showed that the lowest error gave 8 period moving average. The forecast mathematically is computed as follows:

where F(t+1) -forecast in room demand in period t+1, x – is the number of rooms sold in period i, N-the number of past periods (Phumchusri and Mongkolkul, 2012). Simple moving average is simple, fast to compute and respond more quickly to shifts in demand when N period is small. However this method has two major disadvantages. Firstly, it assumes that most recent observations are better predictors than older data. Secondly, when data exhibits upward or downward trend, method will be constantly overforecast or underforcast. In order to cope with such trends Talluri and Van Ryzin (2004) recommend using double or triple moving average.

The application of this method on our data set is available here: Simple Moving Average

Example

In our application of this forecasting method enabled to achieve MAPE of 4%, what is a very good example. However, as it was mentioned before, this method is a poor predictor when the demand is more unstable. The following graph shows such a situation, where MAPE amounted to 60% (in model 2 – forecasted Values_1: 2 periods) and 55% (in model 8 – forecasted values_2: 8 periods).

References:

1. Phumchusri, D., Mongkolkul, J. (2012) Hotel Room Demand via Observed Reservation Information. Proceedings of the Asia Pacific Industrial Engineering & Management Systems Conference 2012, pp. 1978-1985
2. Talluri, K. and Van Ryzin, G. (2004) The theory and practice of revenue management. Boston, Kluwer Academic Publishers.
3. Weatherford, L.R. & Kimes, S.E. (2003). A comparison of forecasting methods for hotel revenue management. International  Journal of Forecasting, vol. 19, no. 3, pp. 401-415.

Share