Table 1. Normalized results for the classification results
obtained with the neural network algorithm
for the classification using 5% of data for training.
Reference categories
Classification
categories Com Soybeans Forest Pasture Bare Soil River
Com 0.9465 0.0026 0.0317 0.0125 0.0011 0.0055
Soybeans 0.0044 0.9275 0.0080 0.0454 0.0025 0.0123
Forest 0.0065 0.0442 0.9389 0.0102 0.0001 0.0002
Pasture 0.0012 0.0221 0.0006 0.8970 0.0789 0.0002
Bare Soil 0.0297 0.0009 0.0010 0.0410 0.9271 0.0003
River 0.0002 0.0001 0.0078 0.0001 0.0000 0.9918
Table 2. Normalized results for the classification results
obtained with the neural network algorithm
for the classification using 10% of data for training.
Reference categories
Classification
categories Com Soybeans Forest Pasture Bare Soil River
Com 0.9528 0.0080 0.0122 0.0200 0.0006 0.0064
Soybeans 0.0035 0.9121 0.0321 0.0412 0.0001 0.0111
Forest 0.0185 0.0169 0.9408 0.0234 0.0001 0.0003
Pasture 0.0007 0.0414 0.0029 0.9145 0.0404 0.0001
Bare Soil 0.0019 0.0195 0.0005 0.0008 0.9770 0.0003
River 0.0082 0.0001 0.0035 0.0001 0.0000 0.9880
Table 3. Normalized results for the classification results
obtained with the neural network algorithm
for the classification using 15% of data for training.
Reference categories
Classification
categories Com Soybeans Forest Pasture Bare Soil River
Com 0.9607 0.0059 0.0072 0.0182 0.0012 0.0068
Soybeans 0.0031 0.9619 0.0272 0.0075 0.0001 0.0002
Forest 0.0146 0.0064 0.9561 0.0222 0.0003 0.0004
Pasture 0.0005 0.0160 0.0010 0.9116 0.0586 0.0123
Bare Soil 0.0009 0.0006 0.0052 0.0445 0.9486 0.0002
River 0.0077 0.0001 0.0000 0.0001 0.0000 0.9920
Table 4. Normalized results for the classification results
obtained with the neural network algorithm
for the classification using 20% of data for training.
Reference categories
Classification
categories Com Soybeans Forest Pasture Bare Soil River
Corn 0.9692 0.0043 0.0108 0.0082 0.0002 0.0073
Soybeans 0.0026 0.9570 0.0066 0.0212 0.0028 0.0098
Forest 0.0087 0.0016 0.9875 0.0013 0.0005 0.0004
Pasture 0.0004 0.0226 0.0009 0.9505 0.0255 0.0001
Bare Soil 0.0067 0.0062 0.0001 0.0122 0.9748 0.0001
River 0.0001 0.0001 0.0026 0.0001 0.0000 0.9970
532
Table 5. Performance summary of
the neural network classifiers corresponding
to the trainings with 5%, 10%, 15%, and 20% TM data.
Classification Classifiers
categories
NN-5% NN-10% NN-15% NN-20%
Com 0.9465 0.9528 0.9607 0.9692
Soybeans 0.9275 0.9121 0.9619 0.9570
Forest 0.9389 0.9408 0.9561 0.9875
Pasture 0.8970 0.9145 0.9116 0.9505
Bare Soil 0.9271 0.9770 0.9486 0.9748
River 0.9918 0.9880 0.9920 0.9970
Table 6. SAS output from the multiple comparisons
of the classifiers corresponding to the trainings
with 5%, 10%, 15%, and 20% TM data.
General Linear Models Procedure
Tukey’s Studentized Range (HSD) Test for variable: Y
NOTE: This test controls the type I experimentwise error rate,
but generally has a higher type II error rate than REGWQ.
Alpha- 0.05 df= 14 MSE= 0.00017
Critical Value of Studentized Range=4.111
Minimum Significant Difference= 0.0219
Means with the same letter are not significantly different.
Tukey Grouping Mean N METHOD
A 0.9727 6 NN-20%
A
B A 0.9552 6 NN-15%
B
B 0.9475 6 NN-10%
B
B 0.9381 6 NN-5%
60 =
Training rate (seconds/cycle)
NN-5% NN-10%
NN-15%
Figure 1. Training rates of
the four neural network classifiers.