Restricted exponential forgetting in real-time identification
Automatica (Journal of IFAC)
Journal of Computational and Applied Mathematics
| Hi-index | 0.00 |
We propose a multivariate dynamic demand model and also suggest modifications of estimation procedures for real-time model update. The presented estimators account for the presence of incorrect or missing records. Also their numerically stable realizations are explained. The missing data in the market systems is encountered when a subset of products is not on the market for a period of time. Another possible reason for a lack of data is database errors or problems encountered during the data collection. The last cases also can cause incorrect values of the input-output (IO) data. A cross effects determination (CED) procedure is offered, which is used for the multivariate model structure selection. A modified validation criterion, which accounts the lack of data, is explained too. To assess the benefit of the modified procedures we apply a Monte Carlo simulation.