Jurnal Publikasi STMIK Pontianak

Penerapan Metode Jaringan Saraf Tiruan Backpropagation untuk Memprediksi Nilai Ujian Sekolah


Abolishment policy National Examination for the primary level/equivalent during the period of 2013- 2014 have consequences on the growing position of the Examination Schools, particularly in the determination and measurement of competency graduation students. Therefore, as a preparation for the war against it, the author makes an application of artificial neural networks using back propagation method for predicting the value of the Examination Schools elementary school students. This research using case study design, located at SDN 1 Central Singkawang and using experimental methods. A research variable grades of Mathematics and Science subjects as well as the value of the Examination Schools on both these subjects. Methods of design and development using prototyping model. The results showed that the value of the Mean Square Error (MSE), the smallest in Mathematics obtained at 0.5100175 with a combination of parameters in the form of 26,000 training epochs and learning rate of 0.5. In science subjects, the smallest MSE value obtained through a combination of 0.1405143 1,000 epochs of training parameters and the value of learning rate 0.9. The average accuracy rate of the network output obtained for 80.15 %. It can be concluded that backpropagation neural network produced reliable enough to predict school test scores of elementary school students.

Keywords: Artificial Neural Networks, Backpropagation, Prediction, Value of Exam Schools, MSE.

Jurnal Publikasi STMIK Pontianak By SANDY KOSASI