Information Technology Reference
In-Depth Information
genes are shown separately and a dot is used to separate the different ele-
ments; the heads are shown in bold):
NET.d41.d8.+.d12.d8.GOE .d37.d44.d9.d38.d28.d47.d0.d9
d25.d25.d12.d44.d18.NET.ET .d16.d42.d41.d50.d14.d39.d50.d12
-.NET.+.LT.GT.NET.LOE .d20.d46.d16.d40.d48.d34.d41.d4
ET.GOE.d46.d27.d13.d7.LOE .d25.d9.d30.d2.d40.d12.d49.d41
NET.*.d42.d18.d7.+.d20 .d2.d41.d35.d42.d28.d13.d5.d38
*.d24.GOE.d38.d43.d24.d7 .d24.d37.d2.d45.d49.d24.d47.d34
d38.d28.d0.GOE.d10.GOE.d27 .d49.d40.d0.d9.d44.d45.d27.d19
GOE.d21.+.GOE.NET.LOE.d31 .d47.d42.d5.d27.d22.d49.d33.d25
+.d2.d23.d32.LT.NET.d44 .d40.d31.d22.d21.d3.d15.d38.d1
d17.d20.LT.ET.d37.-.d34 .d42.d11.d48.d19.d18.d16.d23.d3 (4.10a)
It has a fitness of 310 evaluated against the training set of 345 fitness cases
and a fitness of 151 evaluated against the testing set of 172 samples. This
corresponds to a testing set classification error of 12.209% and a classifica-
tion accuracy of 87.791%, almost identical to the classification error and
classification accuracy obtained in the training set, which is indicative of the
excellent generalizing capabilities of the evolved model. Below is shown the
fully expressed individual translated into a C++ function automatically gener-
ated with Gepsoft APS:
int apsModel(double d[])
{
const double ROUNDING_THRESHOLD = 0.5;
double dblTemp = 0.0;
dblTemp = (d[41]!=d[8]?1:0);
dblTemp += d[25];
dblTemp += (((d[20]<d[46]?1:0)!=(d[16]>d[40]?1:0)?1:0)-
((d[48]!=d[34]?1:0)+(d[41]<=d[4]?1:0)));
dblTemp += ((d[27]>=d[13]?1:0)==d[46]?1:0);
dblTemp += ((d[18]*d[7])!=d[42]?1:0);
dblTemp += (d[24]*(d[38]>=d[43]?1:0));
dblTemp += d[38];
dblTemp += (d[21]>=(((d[5]<=d[27]?1:0)>=d[31]?1:0)+
(d[47]!=d[42]?1:0))?1:0);
dblTemp += (d[2]+d[23]);
dblTemp += d[17];
return (dblTemp >= ROUNDING_THRESHOLD ? 1:0);
} (4.10b)
Note that not all the attributes integrate the model evolved by GEP, which
indicates that they are apparently irrelevant to the decision at hand. In fact,
of the 51 attributes only 25 are used in this extremely accurate model. This is
Search WWH ::




Custom Search