|
|
Reply from Mr. Dodi Zulherman
2020.11.18 14:21:51
Dear Mr. Chalidia Nurin Hamdani
Thank you for the question, the results of several experiments that have been
tested in this study have a poor level of accuracy. We have found out that the error
value obtained during classification is very high, where the highest error value is
8235.7 during the 3745 iterations. But, there are some times where the error
increases and suddenly decreases again. The smallest error value obtained is
110.2192 in the 17622 iterations. Based on the error graph concept, it can be
concluded that the classification system used by the researcher is underfitting. This
can be caused by several things, including the pre-processing process that can
change the value of the dataset, as well as the extraction of features that have not
yet obtained the characteristics of the dataset, so it is difficult to distinguish each
class.
I hope this can answer your question.
|
