|< Day Day Up >|| |
An application of the method of combination of forecasts has been presented and the value of the improvement in the forecasting performance assessed. The usual approach in practical inventory management is to evaluate alternative forecasting methods over a sample of items and then select the one that gives the lowest errors for a majority of the items in the sample to use for all items being stocked. In the example considered in this paper, the single forecasting method to use would have been the Holt's method if the results of the statistical analysis on seasonality had been accepted or the Holt-Winters if seasonal factors had been imposed contrary to the analysis. It is not part of this research to consider the merits of seasonal versus nonseasonal forecasting models. The theme is whether there is likely to be benefit to practical inventory management from combining values from different demand forecasting models together. The results show that combining different forecasts will lead to significant improvements in demand forecasting performance and savings of around 10% in the amount of safety stock that will need to be carried.
There appears to be no obvious advantage in trying to find the best individual forecasting methods for each item. It is only necessary to determine which is the most common method identified for each forecasting group over a sample of items. This reduces the workload on the bank in running its inventory management system. Updating weights every month is not necessary for the bank; therefore the bank could use the Fixed Weights method, reestimating the weights once a year to reduce workloads. However, it seems advisable to determine the best individual set of weights for each time series. Using a common set of weights gives a much better reduction in the sum of root mean squared errors. Whenever combination of forecasts is not suggested by the method of "optimal" weights, a simple average method seems appropriate, as indicated by the performance of the "hybrid" method. Obviously, a reduction of forecasting errors for the printed forms can save inventory holding costs for the bank, as the safety or buffer stock is directly proportional to these errors.
The above observations from our study seem to give us some insight in the process of data mining. There are a number of well-known methods in data mining such as clustering, classification, decision trees, neural networks, etc. Finding a good individual method from our tool kit to handle the data is clearly an important initial step in data mining. Then we should always bear in mind the power in combining the individual methods. In this study, we found two kinds of direction to do the combination. The first one is basically a direct combination of the individual methods, such as simple average or quadratic programming. The other one is to classify our data first and then apply the hybrid paradigm. Classification is always an important aspect of data mining and our study probably shed some new light on this. The next important message is that if we are dealing with large data sets, then it is not very worth while to find the "best" individual method. Obviously there may not be any best individual at all. A viable alternative is to find several sensible individual methods and then combine them as the final method. This approach will usually relieve much of our effort in finding the best individual method, as justified by the law of diminishing return.
|< Day Day Up >|| |