Categorias
Demanda Previsão Qualidade

Uso de gráficos de controle para avaliar previsões

Em 2009, no Seminário “The 29th Annual International Symposium on Forecasting” em Hong Kong, o trabalho Demand management and inventory control through forecast error monitoring using the combined Shewhar-CUSUM control charts foi apresentado.

Autores: Robert Wayne Samohyl e Leandro Callegari Coelho.

Resumo em inglês: Optimal inventory size is essential  for maintaining the competitive edge. Demand management and forecasting are a vital part of the calculation due to the fact that forecast errors and the size of inventories are correlated. Errors in forecasting should be monitored for the adequacy of the underlying model. In statistical process control, a monitoring scheme that has received much attention recently in the industrial setting but not for forecasting error is the combined Shewhart-CUSUM control chart. Combining the two charts with proper attention paid to the probability of  false alarms is an important part of the new scheme. The CUSUM chart will  detect small changes in the error distribution while the Shewhart chart will detect large alterations. The paper shows that the Shewhart-CUSUM control chart works very well for accompanying the size and movement of inventories. When errors are large, alarms are sounded and re-estimation of the underlying models is called for. A pratical example of the procedure is illustrated which demonstrates the benefits of making forecasts with automatic exponential smoothing models whose forecasting errors are monitored by the  scheme suggested here. The result is that inventories are kept to a minimum.

Categorias
Demanda Previsão Qualidade

Avaliação estatística de modelos de previsão.

Em 2009, no Congresso “Optimization Days” em Montréal no Canadá, o trabalho Evaluation of Exponential Smoothing Models for Demand Forecasting with Combined Shewhart-CUSUM Control Charts foi apresentado. O congresso aconteceu no HEC Montréal (www.hec.ca).

Autores: Leandro Callegari Coelho e Robert Wayne Samohyl.

Veja detalhes da apresentação clicando aqui.

Resumo em inglês: We consider the problem of modeling the demand using exponential smoothing forecasts methods and to evaluate and control it using combined Shewhart-CUSUM control charts, with control limits calculated by simulation for this specific application. This control chart acts as an alarm for the need of re-estimation of the exponential smoothing model and its components.