Information Technology Reference
In-Depth Information
models confers
KBIF
with a great expansibility and flexibility in rule
definition.
Model One: Frame Model
KBIF
. The knowledge (mainly many kinds of
restriction factors) in
is coded in terms of rules and the domain experts
can define the impact factors flexibly. These factors form a consistent system
through inter-factor communications and reach the goal collaboratively.
The atom set is constructed as follows: First of all, collect the restricting
factors in the formation of central fishing grounds, and divide them as the
primitive elements of prediction rules. Each element is an atom of prediction
rules, while all these elements form the atom set, i.e. the frame. The atom set
utilizes knowledge representation principle of
KBIF
to express all influence
factors that can be expressed in the prediction algorithm, and every factor can
be treated as a slot in the frame. The frame is named as “The rule set of
revision system” and layered into the following three layers: The station data,
flow rate of the Changjiang River, and flow divisions. The station data
include: Wind speed, wind direction, precipitation, temperature, atmospheric
pressure, time, water temperature, salt degree, name of station, etc. The flow
rate comprises time, month average. The flow divisions include: surface
temperature, surface temperature gradient, surface salt degree, surface salt
gradient, vertical water temperature gradient, vertical salt gradient, ground
floor salt degree, ground floor salt d gradient, time, water temperature of
ground floor and water temperature of ground floor. Atom set can be
extended along with the enhancement of domain experts' understanding and
meet the expanding requirement on prediction of central fishing grounds.
Model Two: Blackboard structure. We have developed the subsystem of
blackboard in a broad sense on the basis of
KBIF
, which plays an important
role in the decision process. With this subsystem, the users can divide the
solution space into sub-structures, i.e., rule sets by homogeneity and
heterogeneity cut arbitrarily.
We have defined the notion of virtual blackboard, which can be instanced
to blackboard systems in accordance with the predication characteristics of
fishery resources. It combines the knowledge sources, i.e., atom sets, rule
generation set, and concrete application of fuzzy reasoning. The architecture
of KBIF is a miniature blackboard, including a group of atom sets, a
miniature blackboard and a controller for rule production. The architecture
offers an effective combination of a knowledge layer and an experience
reasoning layer. Atom sets are coded in the system implicitly, corresponding
KBIF
