61
IV. RESULT AND DISCUSSION
4.1. Performances of GIS Raster and ANN Method 4.1.1. Training of ANN
Data set for training and validation were composed from a combined raster multilayer database. The database contains ten single layers, which were combined
using ‘Combine’ functionality of ArcGIS9.0. Number of training data set recorder were 319, which were derived from raster multilayer database, consisting of 58 of
S1, 222 of S2, 24 of S3 and 15 of N. The 319 training datasets of each suitability were obtained from the score of each layer, so that the level of suitability refers to as
in Table 3.2. The suitability of the training dataset being used was formed into binary as in Table 3.4.
Using the logistic constant of 0.2 and momentum constant β of 0.3, the
number of iteration obtained was 377, the result of training process can be seen in Figure 4.1.
20 40
60 80
100 120
1 50
100 150
200 250
300 350
Iteration Epoch Accuracy Level
Figure 4.1. The relationship between accuracy and iteration epoch number
The result indicates that using the logistic constant and learning rate the iteration process will be stabled at 377 on which the accuracy achieve 97 as shown
in Table 4.1. This accuracy value indicates that the data training set are highly
62 consistent. The S1, S3, and N were achieved highest accuracies due to the input
pattern are highly of representative to produce the output suitability. Also, it may occur due to the number of input in training data set are suitable to the number of
output. In contrast, for S3 was achieved at an accuracy of 96. It was caused by the input pattern that was less representative to produce output or it can be the number of
input of training data set that were less suitable to the number of output. However, the S3 accuracy constitutes a high accomplishment.
Table 4.1. Accuracy of ANN for training dataset processing
No. Output Label Number of
Training Dataset Number of Valid
Prediction Accuracy
1. S1
58 58
100 2.
S2 222
214 96
3. S3
24 24
100 4.
N 15
15 100
Total 319
310 97
4.1.2. Validation
Validation of the process was conducted using validation data set that were also resulted from combined raster multilayer database. The number validation dataset
was 233, and was dedicated for testing the ANN performance. Validation was done to check and verify the consistency of system. Validation process was conducted as
if the computation running the training data set. So, the number of logistics constant and learning rate were similar to the training process without iteration process.
Computation using training data set was to determine the weight of the ANN. Weight value was then used as input for validation data set to produce overall
accuracy. Overall accuracy represents the performance of the system that being assessed. Overall accuracy values represent the steps to test the validity whether the
system run well or not. If the result were not satisfactory, for instance, it will be recalculated back to the learning step. Meanwhile, if the accuracy were good, then go
to the map prediction to map the suitability see Table 4.2.
63 The overall accuracy of validation dataset is 96, and this indicates that the
performance system was good. This accuracy can be accepted to the prediction of suitability class processing. As discussed in previously, lower accuracy was achieved
due to less representation between the input and output pattern. Also, it may occur because the number of training dataset was less suitable. However, the overall
accuracy of 95 was achieved, and this indicates that the system has met the requirement to be continued for prediction.
Table 4.2. Accuracy of ANN for validation dataset processing
No. Output Label Number of
Training Dataset Number of Valid
Prediction Accuracy
1. S1
25 25
100 2.
S2 83
77 97
3. S3
23 19
83 4.
N 2
2 100
Total 233
216 96
4.2. Land Suitability Map of GIS Raster and ANN