Using CBR as Design Methodology for Developing Adaptable Decision Support Systems - Efficient Decision Support Systems: Practice and Challenges from Current to Future - page 136

Attribute Type Description

Location AlphanumericGeographical location of the sensor buoy.

Date dd-mm-yyyy Date on which the measure was made.

Time hh-mm-ss Time on which the measure was made.

Temperature Numeric Water temperature (cent. degrees) at different depths.

Oxygen Numeric Oxygen concentration (milliliters/liter) at different depths.

PH Numeric acid/based scale.

Transmitance Numeric Fraction (percentage) of sun light that passes through the

sea-water.

Fluorescence Numeric Sea-water fluorescence (percentage) at different depths.

Cloud index Numeric Cloud measurement derivate from a geostationary satellite.

Recount of
diatoms

Numeric

Algae concentration (in cell/liter) at different depths.

Pseudo-nitzschia
spp

Numeric Diatom concentration (in cell/liter) at different depths

causing harmful algae blooms.

Pseudo-nitzschia
spp
(future)

Numeric Diatom concentration (in cell/liter) to be predicted.

Table 3. Oceanographic parameters and physical characteristics of the water mass
comprising a case in the FSfRT system

Fig. 8. Life cycle of the FSfRT system for oceanographic prediction
In the revision phase, the initial solution proposed by the RBF neural network is modified
according to the responses of the four fuzzy revision subsystems. Each revision subsystem
has been created from the RBF network using neurofuzzy techniques (Jin & Sendhoff, 2003).
For each class of the GCS neural network a vector of four values is maintained (see Figure