Monitoring Tools for the Detection of Abnormal Forecasts (Variant 1)
We developed heuristic plausibility checks for a forecasting system, which are executed regularly and inform the responsible employees about conspicuous forecasts with automatically generated e-mails.
The basis for the definition of the heuristic plausibility checks was the experience of the employees of VIA Software from a previous project, the subject of which was the technical support of the operation of a forecasting system. The heuristics applied in this project were formalized and suitably parameterized, which – compared to a manual check – enabled an automated heuristic evaluation for a large number of forecasts. The anomalies were also automatically assigned a “severity” based on an estimate of the potential monetary impact of possible forecast errors, which was used to prioritize downstream manual analyses.
Furthermore, we extended the automated analyses with additional heuristic checks and statistics defined by the customer and set up regular reporting of anomalies via e-mails to the respective market experts.