Information Technology Reference
In-Depth Information
is quite difficult too, in that the position and size of fishing ground are space
data. Meanwhile, the information of sea state involves about 600 pieces of
space and un-space attributes, methods such as the decision tree, regression,
etc. are not very suitable. On the basis of the above consideration, we adopt
CBR, a kind of Lazy Learning (Aha, 1997).
A rational, consistent case representation is essential in the process of
case library building. A prominent question in a CBR system is how to
combine situation, solution and outcome to form a case representation. In our
system, we employ the Object Orientated Technologies to express the sea
state and fishing condition, and every instance of a class is captured by a
record in database and created dynamically during the system evolvement.
Through the proper pre-treatment, the initial data are turned into forms
suitable for machine learning. The primitive records take the form of time,
place, output, manner, etc. (records from some fishery company even have no
output records), upon which we must compute the central fishing grounds.
Meanwhile, for simplicity, we neglect fishing grounds' form and only take
their sizes into consideration. We classify these primitive records into groups
in terms of time and manner, and can do some reduction in some clustering
ways to find central fishing grounds, which can be revised in the following
phase through human-computer interaction:
(1) Combine the adjacent central fishing grounds around until there are no
adjacent ones;
(2) Eliminate the central fishing ground with output lower than a preset
threshold;
(3) If the number of central fishing grounds is still big after the above
processes, eliminate the fishing grounds with smaller size and modest
output;
(4) Combine the remaining fishing grounds to form the central fishing ground:
the position of the central fishing ground is their geometric centre, its size
is the number of the component fishing grounds and the output is their
accumulation.
(5) alize the result of Step 4, and users do some revision according to the
specific situations;
(6) the result in database.
