IJCCS (Indonesian Journal of Computing and Cybernetics Systems)
Not a member yet
480 research outputs found
Sort by
The Comparison of ReliefF and C.45 for Feature Selection on Heart Disease Classification Using Backpropagation
Full text linkOne of the datasets used to classify heart disease is UCI dataset. unfortunately, the dataset contains missing data. Backpropagation is an easy and fast method, but it is very dependent on input data so if there is missing data, it can reduce the performance of the backpropagation. One of the techniques used to handle missing data is feature selection. This study compares ReliefF and C4.5 algorithm in feature selection. The purpose of the study is to find way in overcoming missing data by feature selection to improve backpropagation performance in the heart disease classification. The results of these algorithms are applied to the classification by Backpropagation. The results will be measured based on accuracy, precision, and recall. The performance results of the ReliefF and Backpropagation are above 82%. The performance results of of C4.5 and backpropagation are 80.54% on average for accuracy, recall and precision. Based on the results it can be concluded the ReliefF gives better performance on backpropagation than C4.5. ReliefF is also able to handle missing data by performing feature selection to improve the performance of the backpropagation method for heart disease classification compared to C4.5. Although the C4.5 algorithm is able to provide increased performance on backpropagation, C4.5 is not appropriate to be used as a feature selection method for handling missing data
Optimizing ODP Device Placement on FTTH Network Using Genetic Algorithms
Full text linkCurrently the problem of Optical Distribution Point (ODP) infrastructure is important in fiber to the home (FTTH) network access because ODP infrastructure development is no longer dependent on demand, so placing ODP manually without a systematic method can cause an increase in the value of optical fiber attenuation. on the length of the cable and cause the cable distribution to be irregular. This study aims to optimize the placement of ODP devices in PT BCV's FTTH network by using the Traveling Salesman Problem (TSP) scheme with the genetic algorithm (GA) approach and using hybrid GA, testing is carried out using Matlab software. Testing with development using Hybrid GA gets the best path with a fitness value of 28.6457 and a computation time of 89.93 seconds
Aspect-Based Sentiment Analysis in Bromo Tengger Semeru National Park Indonesia Based on Google Maps User Reviews
Full text linkTechnology can influence and shape a person's behavior patterns when planning tours, traveling, and after traveling. Visitors' reviews can be used as evaluation material to improve the quality of tourist destinations and become a determining factor for other tourists to visit or revisit the destinations. The process of utilizing these reviews can be done by assessing the aspects of tourist destinations based on reviews from visitors. This study aims to conduct an aspect-based sentiment analysis on one of the tourist destinations in Indonesia, namely Bromo Tengger Semeru National Park, based on reviews of Google Maps users. The aspects consist of attractions, facilities, access, and price. The sentiment classification model used is a machine learning model consisting of SVM, Complement Naïve Bayes, Logistic Regression, and transfer learning from pre-trained BERT, IndoBERT, and mBERT. Based on the experimental results, transfer learning from the IndoBERT model achieved the best performance with accuracy and F1-Score of 91.48% and 71.56%, respectively. In addition, among the machine learning models used, the SVM model gives the best results with an accuracy of 89.16% and an F1-Score of 62.23%
World Cup 2022 Knockout Stage Prediction Using Poisson Distribution Model
Full text linkFootball is one of the most popular sports in the world. The popularity makes every topic related to football interesting, for instance, the FIFA World Cup winner prediction. This topic is not only for casual discussion but could be a practical decision support for coaching staff to rate the team’s readiness. Most prediction methods use large match datasets. Since every national team has a different squad for every world cup and the FIFA World Cup is held every four years, the usage of a large match dataset is irrelevant. Therefore, there is a need for a prediction method based on the relevant data. We applied the Poisson distribution model for predicting the FIFA World Cup 2022 knockout stage match results. We calculate the probability of winning and losing based on their average goal scores and goal conceded and evaluate the difference by the actual result using de Finetti distance. The successful prediction is 8 out of 15 matches, with six inside the round of 16 games. This prediction model is also a brief example to overcome prediction problem using limited dataset. Thus, the new data attributes need to reformulate Poisson’s lambda. Further studies need to add the 3-4 prior world cup matches data to increase the acceptance of prediction
Smart GreenGrocer: Automatic Vegetable Type Classification Using the CNN Algorithm
Full text linkIn the food industry, separating vegetables is done by visually trained professionals. However, because it takes plenty of time to sort a large number of different types of vegetables, human errors might arise at any time, and using human resources is not always effective. Thus, automation is needed to minimize process time and errors. Computer vision helps reduce the need for human resources by automatizing the classification. Vegetables come in various colors and shapes; thus, vegetable classification becomes a challenging multiclass classification due to intraspecies variety and interspecies similarity of these main distinguishing characteristics. Consequently, much research is made to automatically discover effective methods to group each type of vegetable using computers. To answer this challenge, we proposed a solution utilizing deep learning with a Convolutional Neural Network (CNN) to perform multi-label classification on some types of vegetables. We experimented with the modification of batch size and optimizer type. In the training process, the learning rate is 0.01, and it adapts on arrival in the local minimum for result optimization. This classification is performed on 15 types of vegetables and produces 98.1% accuracy on testing data with 25 minutes and 45 seconds of training time
Flower Pollination Inspired Algorithm on Exchange Rates Prediction Case
Full text linkThe flower pollination algorithm is a bio-inspired system that adapts a similar process to a genetic algorithm that aims for optimization problems. In this research, we examine the utilization of the flower pollination algorithm in linear regression for currency exchange cases. Each solution represents the regression coefficients. The population size for the solutions and the switching probability between global pollination and local pollination is experimented with in this research. The result shows that the final solution is obtained using a larger population and higher switch probability. Furthermore, our research finds that the increasing population size leads to considerable running time, where the probability of global pollination just slightly increases the running tim
C Source code Obfuscation using Hash Function and Encryption Algorithm
Full text linkObfuscation is a technique for transforming program code into a different form that is more difficult to understand. Several obfuscation methods are used to obfuscate source code, including dead code insertion, code transposition, and string encryption. In this research, the development of an obfuscator that can work on C language source code uses the code transposition method, namely randomizing the arrangement of lines of code with a hash function and then using the DES encryption algorithm to hide the parameters of the hash function so that it is increasingly difficult to find the original format. This obfuscator is specifically used to maintain the security of source code in C language from plagiarism and piracy. In order to evaluate this obfuscator, nine respondents who understand the C programming language were asked to deobfuscate the obfuscated source code manually. Then the percentage of correctness and the average time needed to perform the manual deobfuscation are observed. The evaluation results show that the obfuscator effectively maintains security and complicates the source code analysis
Deep Learning Approaches for Nusantara Scripts Optical Character Recognition
Full text linkThe number of speakers of regional languages who are able to read and to write traditional scripts in Indonesia is decreasing. If left unaddressed, this will lead to the extinction of Nusantara scripts and it is not impossible that their reading methods will be forgotten in the future. To anticipate this, this study aims to preserve the knowledge of reading ancient scripts by developing a Deep Learning model that can read document images written using one of the 10 Nusantara scripts we have collected: Bali, Batak, Bugis, Javanese, Kawi, Kerinci, Lampung, Pallava, Rejang, and Sundanese. While previous studies have made efforts to read traditional Nusantara scripts using various Machine Learning and Convolutional Neural Network algorithms, they have primarily focused on specific scripts and lacked an integrated approach from script type recognition to character recognition. This study is the first to comprehensively address the entire range of Nusantara scripts, encompassing script type detection and character recognition. Convolutional Neural Network, ConvMixer, and Visual Transformer models were utilized and their respective performances were compared. The results demonstrate that our models achieved 96% accuracy in classifying Nusantara script types, with character recognition accuracy ranging from 93% to approximately 100% across the ten scripts
Mahalanobis Fuzzy C-Means Clustering with Spatial Information for Image Segmentation
Full text linkA fuzzy C-Means segmentation algorithm can be implemented in an image segmentationbased on the Mahalanobis distance; However, this method only needs to consider the colorspace situation, not the neighborhood system of the image. It was an effective edge detectionprocess unwell performed and generated less accuracy in segmentation results. In this article,we propose a new method for image segmentation with Mahalanobis fuzzy C-means Spatialinformation (MFCMS). The proposed method combines feature space and images of theinformation of the neighborhood (spatial information) to improve the accuracy of the result ofsegmentation on the image. The MFCMS consists of two steps, the histogram threshold modulefor the first step and the MFCMS module for the second step. The Histogram Threshold moduleis used to get the MFCMS initialization conditions for the cluster centroid and the number ofcentroids. Test results show that this method provides better segmentation performance thanclassification errors (ME) and relative foreground area errors (RAE) of 1.61 and 3.48,respectively
Concrete Subsurface Crack Detection Using Thermal Imaging in a Deep Neural Network
Full text linkImpact actions, such as a zone directly affected by conflict and warfare, can negatively impact the structural integrity of concrete structures. Even indirect impact actions can make structures unsafe, creating subsurface defects in concrete. However, the result of indirect impact actions is often undetected because of the time required and expert knowledge needed to assess the structure. Yet, there are no techniques currently available to assess the usability and the safety of a concrete structure rapidly and with no expert knowledge.. This paper presents a combination of thermal imaging and artificial intelligence (AI) to enable a novel, contactless, autonomous, and fast technique for detecting hidden defects in concrete structures. In this paper, a ResNet50 model was trained on simulated data of subsurface defected and defect-free concrete blocks to test if it is possible to classify between the two. The model developed achieved a validation accuracy of 99.93%. Because of the success of this model, a laboratory experiment was conducted by compressing concrete blocks and recording the process using a thermal camera to create a dataset of concrete blocks with and without subsurface cracks. This dataset was used to train a new model with the same architecture and hyper-parameters as the initial model and achieved a validation accuracy of 100%. This investigation proves it is possible for AI to detect subsurface cracks and hidden defects by classifying the thermal images of concrete surfaces