| Penulis | Agung Toto Wibowo, Husnul Khotimah, Warih Maharani |
| Abstract | “Artificial Neural Network (ANN) with backpropagation algorithm is widely used for classification or prediction. That is because the basic capabilities of the ANN in generalizing the given input information like a human being that can recognize specific patterns of an object based on the general pattern that has been learned and saved in the memory. A System that capable of predicting the value of specific subjects is required by students during the next semester registration. Commonly they do not know what subjects that should have been taken because the problem of interests and capabilities of the student. ANN, which can predict the value of certain subjects from a student can be used in addressing these problems. The architecture cconfigurations of an ANN determine considerably the quality of its network. Although the method that usually used is try and error, the determination of an ANN can be done based on the case that being tested. ANN Multi Layer Perceptron (MLP) with back propagation algorithm is often used for the case which the output is nonlinear. In this research, ANN architectures were tested to predict the twelve subjects in six semesters. Each subject has a grade index of A to E, so this case can be classified as a multiclass prediction. Performance will be measured using the MSE in the form of normalization and denormalization. The input is preprocessed by reducting the attribute conducted using Information Gain (IG) and then took five attributes for each subjects are needed with the highest score of IG. MSE that was obtained from the experiment are quite diverse in the twelfth such subjects. This is due to input quality aspects different. Best value for normalization and denormalization are 0.024 and 1.044. And the Average MSE value for the entire subject are 0.064 for normalization and 1.895 for denormalization.” |
| Nama Konferensi | International Conferrence on Tellecomunication 2009 |
| Penyelenggara | IT Telkom |
| Tanggal |
| Link Konferensi |
| ISSN / ISBN |
| Link Paper / Download | https://www.dropbox.com/s/ksplnhzpq1w1066/Multi%20Layer%20Perceptron%20Classifier%20Analysis%20To%20Predict%20Multiclass%20In%20It%20Telkom%20Student%E2%80%99s%20Grade%20Study%20Case.pdf?dl=0 |