PUTRA, BIONDI BAGASTA WIKO (2019) PENERAPAN ALGORITMA JARINGAN SARAF TIRUAN METODE BACKPROPAGATION UNTUK MEMPREDIKSI JUMLAH NILAI EKSPOR DI PROVINSI NTB. S1 thesis, Universitas Mataram.
![[img]](http://eprints.unram.ac.id/style/images/fileicons/text.png) |
Text
Laporan_Skripsi_Biondi_F1D014018.pdf Restricted to Repository staff only Download (2MB) |
Abstract
This paper presents the application of the Backpropagation method of the Artificial Neural Network algorithm in the case study to predict the amount of export value in NTB province. This forecasting process uses two scenarios namely forecasting the total value of total exports in NTB province and forecasting the amount of export value based on commodity. The activation functions used are Sigmoid Binary and ReLU - Linear. Based on the results of the system tests that have been carried out, the Sigmoid Binary activation function shows that the best network architecture is 12-4-1, the best learning rate is 0,3 and the best error limit is 0,0015, which in the test phase produces an MSE value of 0,0161 and the MAPE value of 30,53%, while the ReLU-Linear activation function shows that the best network architecture is 12- 5-1, the best learning rate is 0,1 and the best error limit is 0,0012, which is at the test phase produce MSE value of 0,0309 and the MAPE value of 53,04%. Because the MAPE value generated in the system testing using each activation function does not have a value of less than 20%, the use of Backpropagation in the study is not suitable.
| Item Type: | Thesis (S1) |
|---|---|
| Keywords (Kata Kunci): | ANN, Backpropagation, Value of Exports, System Testing. |
| Subjects: | T Technology > T Technology (General) |
| Divisions: | Fakultas Teknik |
| Depositing User: | Wiwin Kartikawati |
| Date Deposited: | 08 Oct 2019 04:52 |
| Last Modified: | 08 Oct 2019 04:52 |
| URI: | http://eprints.unram.ac.id/id/eprint/14663 |
Actions (login required)
 |
View Item |
