296
J. Bobbin . F. Recknagel
I
I
I
, ,
I
I
L-_ _ ~ _ _ _ -.J ,
, ,
,
I
I
,
I
I
,
I
I
I
I
I
, ,
Figure 15.2. Main steps of the evolutionary algorithm for model induction
15.2.3
Greedy Partitioning Aigorithms
The classification and regression tree (CART) method introduced by Breiman et
al. (1984) is used to compare the performance of the evolutionary method on this
data set with that of a classical machine learning algorithm. The implementation
of the CART algorithm used for this study is that available for the R statistical
package.
5.3
Results
15.3.1
Prediction of Chlorophyll-a
Chlorophyll-a is a measurement of the algal biomass in the lake that changes in
accordance to the growth and succession of different algae species throughout the
years.
The evolutionary algorithm was used to extract patterns in the measured
chemical and physical properties of the lake according to levels of chlorophyll-a.