Article Per Year (5 Year)
p-Index From 2019 - 2024
0.23
P-Index
This Author published in this journals
All Journal Jurnal Teknologi Informasi Cyberku
Heribertus Himawan
Unknown Affiliation
Computer Science & IT Decision Sciences, Operations Research & Management Economics, Econometrics & Finance Industrial & Manufacturing Engineering Languange, Linguistic, Communication & Media Library & Information Science

Published : 5 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 5 Documents
Search

 1 
PREDIKSI KELULUSAN MAHASISWA TEPAT WAKTU BERDASARKAN USIA, JENIS KELAMIN, DAN INDEKS PRESTASI MENGGUNAKAN ALGORITMA DECISION TREE Agus Romadhona; Suprapedi Suprapedi; Heribertus Himawan
Jurnal Teknologi Informasi - Cyberku (JTIC) Vol 13 No 1 (2017): Jurnal Teknologi Informasi CyberKU Vol. 13, no 1
Publisher : Program Pascasarjana Magister Teknik Informatika, Universitas Dian Nuswantoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (275.642 KB)

Abstract

Prediction of the study period in college is needed in determine the accuracy of the students study period according to the specified time so that wisdom of prevention related to the study period is no ton time could be done. This research aims to find patterns to predict the timely graduation of students usingdata mining techniques and models to predict long period of study was Decision tree algorithm C4.5 to compare with ID3 and CHAID algorithms using test data to determine the percentage of precision, recall and accuracy is obtained that the algorithm Decision Tree C4.5 has a better performance compared with other algorithms. From this research it was found that the prediction of the students study period are affected by incoming students age, gender, GPA semesters 1 through 4 semesters GPA and the most influential is the 4th semester GPA of students graduate on time with a value of 0.340 gain of all attributes. Decision tree algorithm C4.5 reaches the highest accuracy on the amount of data 389 with 91.51% accuracy values for k-fold=3, 90.75 for k-fold = 5 and 90.77 with k-fold = 10, While ID3 and CHAID algorithms achieving a low accuracy value. So thus the value accuracy of Decision Tree algorithm C4.5 is better than the ID3 and CHAID algorithm. In this research, training data are used as much as 389. To see better performance in the accuracy of the results of each algorithm, thus for furthermore research the number of data records used training process should be improved.
PROTOTYPE LAMPU LALU LINTAS ADAPTIF BERBASIS MULTI AGENT MANGGUNAKAN LOGIKA FUZZY YANG TERTANAM PADA MICROCONTROLLER Adrin T; Heribertus Himawan; Suprapedi Suprapedi
Jurnal Teknologi Informasi - Cyberku (JTIC) Vol 12 No 2 (2016): Jurnal Teknologi Informasi CyberKU Vol. 12, no 2
Publisher : Program Pascasarjana Magister Teknik Informatika, Universitas Dian Nuswantoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (506.172 KB)

Abstract

Traffic light is a very important tool for urban life in regulating the smooth flow of traffic on the highway. The use of traffic lights are now much more applied using static timing system or a traffic light system which does not know the condition of the crossing are many vehicles or fewer. Timing pattern that is applied from one segment to another segment is set in rotation. In the study conducted to design a traffic light prototype based on multi- agent using fuzzy logic which is planted on the microcontroller. There are two parts and types of microcontrollers are used to design the prototype. The first part, the Slave microcontroller Atmega 32 and the second type, the type of Master mikontroler Atmega 128. Traffic light system which can adapt to the environment made the crossing. If there is a deviation of the queue of vehicles which have very much, then the green light time longer than the deviation that only have a little queue of vehicles. Thus, the traffic light is more adaptive to the dynamic vehicle that will cross the intersection. Traffic lights can also communicate with traffic lights nearest neighbors in both directions. Communication is done through mutual give information about the number of vehicles that left deviation towards each junction nearest neighbors. Results of this study found that the performance of fuzzy logic embedded in the microcontroller can control traffic lights with dynamic adaptive to existing vehicles on Line 1, Line 2 and Line 3. Prototype designed traffic lights represent the environmental conditions that have multiple intersections and each intersection has four traffic lights that can communicate with Multi Agent another intersection nearest neighbors.
PREDIKSI HASIL PENJURUSAN SISWA SEKOLAH MENENGAH ATAS DENGAN MENGGUNAKAN ALGORITMA DECISION TREE C4.5 Imam Sujaj; Purwanto Purwanto; Heribertus Himawan
Jurnal Teknologi Informasi - Cyberku (JTIC) Vol 12 No 1 (2016): Jurnal Teknologi Informasi CyberKU Vol. 12, no 1
Publisher : Program Pascasarjana Magister Teknik Informatika, Universitas Dian Nuswantoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (254.86 KB)

Abstract

The Majors is one of the placement or distribution process in the selection of high school students teaching program. In these majors, students are given the opportunity to choose majors that best matches the characteristics themselves. The accuracy in choosing majors can determine the success of student learning. In contrast, an excellent opportunity for students will be lost due to lack of inaccuracy in determining the majors. In the 2013 curriculum, majors in high school started in class X after being accepted as a student, so the school should really be able to classify students on the correct corresponding majors talents and interests of students. In studies using the C4.5 algorithm to create a predictive model results placement of students because this method has been used a lot in previous studies to predict the various cases problems with good results. This is evident from the results of the C4.5 algorithm generates a classification accuracy of 96.04% value with a precision of 95.96% class, class recall of 95.92% while the value of AUC (Area Under the Curve) of 0948 + / - 0.028 with very good category. It can be concluded that in order to predict the value of the majors C4.5 algorithm produces accuracy that is very good value.
MODEL MULTI-CLASS SVM MENGGUNAKAN STRATEGI 1V1 UNTUK KLASIFIKASI WALL-FOLLOWING ROBOT NAVIGATION DATA Azminuddin I. S. Azis; Vincent Suhartono; Heribertus Himawan
Jurnal Teknologi Informasi - Cyberku (JTIC) Vol 13 No 2 (2017): Jurnal Teknologi Informasi CyberKU Vol. 13, no 2
Publisher : Program Pascasarjana Magister Teknik Informatika, Universitas Dian Nuswantoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1006.645 KB)

Abstract

Manusia memiliki keterbatasan dalam mengerjakan hal-hal yang berat, rumit, cepat, berbahaya, berulang-ulang secara langsung baik itu di kalangan rumah tangga, industri, militer, penelitian, hiburan, dsb. Robot merupakan mesin yang dapat mempermudah pekerjaan dan mengatasi keterbatasan manusia, sedangkan AI dapat membuat robot semakin cerdas. Berbagai macam metode AI telah diusulkan untuk mengatasi salah satu teknik navigasi robot, yaitu wall-following robot navigation, namun masih belum optimal. State of the art dalam klasifikasi wall-following robot navigation data adalah MLP dengan akurasi sebesar 97.59%. Namun akhir-akhir ini, state of the art dalam klasifikasi pattern recognition adalah SVM. Wall- following robot navigation data melibatkan multi-class, non-linear, dan high dimensional problem. 1V1 merupakan strategi terbaik yang dapat diterapkan pada SVM untuk mengatasi multi class problem yang selanjutnya dapat disebut multi-class SVM. Sedangkan untuk mengatasi non-linear dan high dimensional problem, SVM sendiri sudah dimodifikasi dengan memasukkan fungsi Kernel. Dengan demikian, akurasi sebesar 97.59% yang dihasilkan oleh MLP untuk klasifikasi wall-following robot navigation data masih dianggap rendah. Namun model multi-class SVM menggunakan strategi 1V1 untuk klasifikasi wall-following robot navigation data yang diperoleh dengan solusi yang global optimal dan tanpa perlu adanya dimensionality reduction (by PCA/SVD) dalam tingkat akurasi fair classification, yaitu 91.10% < 97.59% yang dihasilkan penelitian sebelumnya menggunakan MLP. Namun secara teoritis, multi-class SVM menggunakan strategi 1V1 lebih cepat dengan waktu proses yang dihasilkan = 10.7505 detik. Dengan demikian, model tersebut dapat mempelajari navigasi robot pengikut dinding tanpa tabrakan (menjaga jarak terhadap dinding dengan baik).
ALGORITMA SUPPORT VECTOR MACHINE UNTUK MEMPREDIKSI NILAI UJIAN NASIONAL Emi Rizky; Purwanto Purwanto; Heribertus Himawan
Jurnal Teknologi Informasi - Cyberku (JTIC) Vol 11 No 2 (2015): Jurnal Teknologi Informasi CyberKU Vol.11 no 2
Publisher : Program Pascasarjana Magister Teknik Informatika, Universitas Dian Nuswantoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (719.689 KB)

Abstract

In order to improve the quality of graduate education through an exam in order to compete in domestic, regional and international levels and therefore require the achievement of national standards through the National Examination (UN). produce test scores that boast with the title and can pass the National Exam, due to lack of graduates when the National Examination become routine issues annually. This problem is felt by students, parents, teachers, educational units and agencies associated with other national exams. By looking at the reasons we need a prediction to predict the value of the UN. Soft computing has several abilities one of which is a technique that can be used to predict the ability of students to acquire the methods of the National Examination Support Vector Machine (SVM) which is a branch of artificial intelligence where the processing system configuration information obtained performance model for the prediction of the National Examination the Root mean squared Error (RMSE) is the best for Indonesian was 0.713 + / - 0.173, English at 0586 + / - 0.066, and Mathematics by 0882 + / - 0188. configuration with predictions using a barometer. k-fold 10, C (cost) of 0.1 with kernel-type radial Indonesian subjects, k-fold 10, C (cost) of 0.3 with radial kernel type for the subjects of English and Mathematics.
Co-Authors Adrin T Agus Romadhona Azminuddin I. S. Azis Emi Rizky Imam Sujaj Purwanto Purwanto Suprapedi Suprapedi Vincent Suhartono