Articles

Found 6 Documents
Search

Biased support vector machine and weighted-smote in handling class imbalance problem Hartono, Hartono; Sitompul, Opim Salim; Tulus, Tulus; Nababan, Erna Budhiarti
International Journal of Advances in Intelligent Informatics Vol 4, No 1 (2018): March 2018
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26555/ijain.v4i1.146

Abstract

Class imbalance occurs when instances in a class are much higher than in other classes. This machine learning major problem can affect the predicted accuracy. Support Vector Machine (SVM) is robust and precise method in handling class imbalance problem but weak in the bias data distribution, Biased Support Vector Machine (BSVM) became popular choice to solve the problem. BSVM provide better control sensitivity yet lack accuracy compared to general SVM. This study proposes the integration of BSVM and SMOTEBoost to handle class imbalance problem. Non Support Vector (NSV) sets from negative samples and Support Vector (SV) sets from positive samples will undergo a Weighted-SMOTE process. The results indicate that implementation of Biased Support Vector Machine and Weighted-SMOTE achieve better accuracy and sensitivity.
Enhancing Performance of Parallel Self-Organizing Map on Large Dataset with Dynamic Parallel and Hyper-Q Sibero, Alexander F.K.; Sitompul, Opim Salim; Nasution, Mahyuddin K.M.
Data Science: Journal of Computing and Applied Informatics Vol 2 No 2 (2018): Data Science: Journal of Computing and Applied Informatics (JoCAI)
Publisher : Talenta Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1212.692 KB) | DOI: 10.32734/jocai.v2.i2-324

Abstract

Self-Organizing Map (SOM) is an unsupervised artificial neural network algorithm. Even though this algorithm is known to be an appealing clustering method,many efforts to improve its performance are still pursued in various research works. In order to gain faster computation time, for instance, running SOM in parallel had been focused in many previous research works. Utilization of the Graphics Processing Unit (GPU) as a parallel calculation engine is also continuously improved. However, total computation time in parallel SOM is still not optimal on processing large dataset. In this research, we propose a combination of Dynamic Parallel and Hyper-Q to further improve the performance of parallel SOM in terms of faster computing time. Dynamic Parallel and Hyper-Q are utilized on the process of calculating distance and searching best-matching unit (BMU), while updating weight and its neighbors are performed using Hyper-Q only. Result of this study indicates an increase in SOM parallel performance up to two times faster compared to those without using Dynamic Parallel and Hyper-Q.
Genetic Algorithms Dynamic Population Size with Cloning in Solving Traveling Salesman Problem Nababan, Erna Budhiarti; Sitompul, Opim Salim; Cancer, Yuni
Data Science: Journal of Computing and Applied Informatics Vol 2 No 2 (2018): Data Science: Journal of Computing and Applied Informatics (JoCAI)
Publisher : Talenta Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1177.754 KB) | DOI: 10.32734/jocai.v2.i2-326

Abstract

Population size of classical genetic algorithm is determined constantly. Its size remains constant over the run. For more complex problems, larger population sizes need to be avoided from early convergence to produce local optimum. Objective of this research is to evaluate population resizing i.e. dynamic population sizing for Genetic Algorithm (GA) using cloning strategy. We compare performance of proposed method and traditional GA employed to Travelling Salesman Problem (TSP) of A280.tsp taken from TSPLIB. Result shown that GA with dynamic population size exceed computational time of traditional GA.
Advertisement billboard detection and geotagging system with inductive transfer learning in deep convolutional neural network Rahmat, Romi Fadillah; Dennis, Dennis; Sitompul, Opim Salim; Purnamawati, Sarah; Budiarto, Rahmat
TELKOMNIKA (Telecommunication Computing Electronics and Control) Vol 17, No 5: October 2019
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1056.819 KB) | DOI: 10.12928/telkomnika.v17i5.11276

Abstract

In this paper, we propose an approach to detect and geotag advertisement billboard in real-time condition. Our approach is using AlexNet’s Deep Convolutional Neural Network (DCNN) as a pre-trained neural network with 1000 categories for image classification. To improve the performance of the pre-trained neural network, we retrain the network by adding more advertisement billboard images using inductive transfer learning approach. Then, we fine-tuned the output layer into advertisement billboard related categories. Furthermore, the detected advertisement billboard images will be geotagged by inserting Exif metadata into the image file. Experimental results show that the approach achieves 92.7% training accuracy for advertisement billboard detection, while for overall testing results it will give 71,86% testing accuracy.
PEMODELAN PERENCANAAN TERINTEGRASI UNTUK RANTAI SUPLAI DAN STOK PENGAMAN MULTI ESELON Hasibuan, Irwitadia; Sitompul, Opim Salim; Lidya, Maya Silvi
JISTech (Journal of Islamic Science and Technology) Vol 4, No 1 (2019)
Publisher : UIN Sumatera Utara Medan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30829/jistech.v4i1.5344

Abstract

Business environment has strong competition from year to year. This is because various changes and uncertainties fill the competition. Most of the changes or uncertainties in the business world are caused by the increasing of consumer bargaining power in business practices. Consumers have high power in determining their requests that must be fulfilled by business people. Changes or uncertainties are most of the main factors that cannot be anticipated when the business world has strong and uncertain competition. These uncertainties require business people to design an appropriate plan in order to minimize costs, especially inventory costs with consumer demand are still fullfilled. In that design plan, business people must be able to optimize the supply chain. In industrial systems, supply chain optimization and its response are strongly influenced by inventories. Inventories and its numbers are important issues in the supply chain that must be integrated with the optimization of the supply chain to manage demand uncertainty and to maintain customer service levels. This study designs an integrated planning model for supply chain and multi-echelon inventory in determining the location and the numbers of inventory in a supply chain with a general configuration with considering the uncertainty of consumer demand or  all things coming from production time
THE FLOWSHOP SCHEDULING MAKESPAN BY THE ACO-GA ALGORITHM Panggabean, Jonas Franky R; Sitompul, Opim Salim; Nababan, Erna Budhiarti
Jurnal Mantik Vol 3 No 4, Feb (2020): Manajemen, Teknologi Informatika dan Komunikasi (Mantik)
Publisher : Institute of Computer Science (IOCS)

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Flow shop scheduling could be a scheduling model where all jobs that are processed flow within the same direction / path. the matter is usually faced if n jobs are processed on m machines, where what must be done first and what allocates jobs on the machine in order that a scheduled production process are obtained. To validate this algorithm a computational test was done employing a dataset of 60 examples from the Taillard Benchmark. HS algorithm with a comparison of two constructive heuristics from the literature, namely the NEH heuristic and stochastic greedy heuristic (SG). The average results obtained for dataset sizes are 20 x 5 to 50 x 10, that the ACO-GA algorithm has smaller makespan compared to the opposite two algorithms, except for large dataset sizes the ACO-GA algorithm has larger makespan compared to the 2 algorithms above with difference of 1.4 units of your time