Information Technology Reference
In-Depth Information
8). This expert's score vector is initialised with a value of (0.25, 0.25, 0.25, 0.25) and represents
the accuracy of each revision subsystem with respect to a class. During revision, the
appropriate expert's score vector is used to ponder the outputs of each fuzzy revision system.
Each vector value is associated with one of the four revision subsystems. For each
forecasting cycle, the value of the importance vector associated to the most accurate revision
subsystem is increased and the other three values are proportionally decreased. This is done
in order to give more relevance to the most accurate revision subsystem.
The revised forecast is then retained temporarily in the forecast database. When the real
value of the concentration of pseudo nitzschia spp is measured, the forecast value for the
variable can then be evaluated, though comparison of the actual and forecast value and the
error obtained. A new case, corresponding to this forecasting operation, is then stored in the
case base. The forecasting error value is also used to update the importance vector
associated with the revision subsystems of the retrieved class.
The FSfRT system was successfully tested using real data collected from years [1992, 2000]
coming from geographical area A0 (42º28.90' N, 8º57.80' W 61 m). Figure 9 shows a
screenshot of the FSfRT interface implemented for oceanographic forecasting.
Fig. 9. Screenshot of the FSfRT system forecasting red tides in the coastal waters of the North
West of the Iberian Peninsula
Search WWH ::




Custom Search