Chapter 15 . Evolved Predictive Rules for Algal Dynamics
303
CART analysis on Kasumigaura
Predicted and actual chlorophyll-a level
... . Ac",",
- _
i §!
:;
i : }J\\~~j\~~A\
' .: ..
'"
i i i I I i I ti I I i I I I I i I I I , I I I n i In'''T,T. r. n . i i I I I i , i i, i i i i , I 11 I , I I i l iI I I I I I I I
o
o .50 300 0 150 300 0 150 300 0 150 300 0 150 300 0 150 300 0 150 300 0 150 300
19~
\985
1987
1988
1989
1990
1991
1992
Predicted and actual chlorophyll-a level using input set
~aJ
-Pf_
i
~ I
j
: "
..
. .
ti
0
50
100
150
200
250
300
350
0
50
100
\50
200
250
300
350
1986
1993
Figure 15.8. Prediction of chlorophyll-a by means of the CART model based on
input set 1 in Table 15.2. Note: Chl-a measured in Ilg/l.
The comparison between the Tables 15.3 and 15.4 and the corresponding Figs.
15.8 and 15.9 shows that the evolutionary algorithm clearly outperforms CART in
each of the 10 runs performed. Even though the differences between the two
algorithms are not significant, the comparison has demonstrated that the
evolutionary algorithm performs at least as good as a greedy partitioning
algorithm like CART.