IJCCS (Indonesian Journal of Computing and Cybernetics Systems)
Not a member yet
480 research outputs found
Sort by
Error Action Recognition on Playing The Erhu Musical Instrument Using Hybrid Classification Method with 3D-CNN and LSTM
Full text linkErhu is a stringed instrument originating from China. In playing this instrument, there are rules on how to position the player's body and hold the instrument correctly. Therefore, a system is needed that can detect every movement of the Erhu player. This study will discuss action recognition on video using the 3DCNN and LSTM methods. The 3D Convolutional Neural Network method is a method that has a CNN base. To improve the ability to capture every information stored in every movement, combining an LSTM layer in the 3D-CNN model is necessary. LSTM is capable of handling the vanishing gradient problem faced by RNN. This research uses RGB video as a dataset, and there are three main parts in preprocessing and feature extraction. The three main parts are the body, erhu pole, and bow. To perform preprocessing and feature extraction, this study uses a body landmark to perform preprocessing and feature extraction on the body segment. In contrast, the erhu and bow segments use the Hough Lines algorithm. Furthermore, for the classification process, we propose two algorithms, namely, traditional algorithm and deep learning algorithm. These two-classification algorithms will produce an error message output from every movement of the erhu player
Applying Data Mining to Classify Customer Satisfaction using C4.5 Algorithm Decision Tree
Full text linkTight business competition demands business actors to make responsive, timely decisions to survive the uncertainty. Food business, especially cafes, has emerged as one of the most popular business types recently. One cafe concept that draws most customers' interest is modern concepts, friendly service, and affordable prices. Finn Coffee is one of the cafes providing a range of foods and beverages, especially coffee-based beverages. Customer satisfaction defines one's feelings when comparing performance. It denotes customer's responses to their satisfied needs. The term satisfaction itself is described as one's happy expression after receiving a quality product with affordable price and satisfying quality. The present study aimed to analyze cafe customer satisfaction using the C4.5 algorithm with predetermined criteria. Customer satisfaction was classified using C4.5. The algorithm displays the level of customer satisfaction based on the customers' response to the Google form distributed by the cafe employees/owner
Identify Reviews of Pedulilindungi Applications using Topic Modeling with Latent Dirichlet Allocation Method
Full text linkThe emergence of Covid-19 in December 2019 has disrupted life worldwide, including Indonesia. The government has made various efforts to control the pandemic, one of which is the development of an application called PeduliLindungi. This app aims to be a reliable tool for the government and the entire community during the pandemic. As a new regulation, the use of PeduliLindungi has prompted numerous reviews assessing its quality and performance. With the app's emergence and growth, various topics have emerged and become trending among the public. These topics were identified through user reviews of the PeduliLindungi app, using the Latent Dirichlet Allocation (LDA) algorithm. The data, consisting of 15,522 reviews, was collected from the Google Play Store and underwent pre-processing, including dictionary and corpus creation, determining the number of topics, and modeling with LDA. The resulting topic modeling process generated the ten most prominent topics. The outcomes were visualized using word clouds and topic distribution graphs, representing the most discussed aspects of the PeduliLindungi app among users. These topics are considered diverse since each issue has no relation or similarity to one another
Fishku Apps: Fishes Freshness Detection Using CNN With MobilenetV2
Full text linkMarine fish are one of the most promising economic commodities for the Indonesian economy. Marine fish will decrease in protein content along with the decreasing level of freshness of the fish that will be consumed. There are still many people who do not know about the classification of fresh and unfresh fish, so we need a system that can classify which fish are fresh and which are not. Previous studies have succeeded in classifying tuna using a convolutional neural network (CNN) algorithm with an accuracy of 90%. In the preprocessing stage of this research, segmentation is carried out, which aims to separate the object to be studied and the background image, then feature extraction is carried out using a color moment, which aims to get the value of the object to be studied. This research was conducted to increase the accuracy value in the freshness classification of tuna and also to add some fish for freshness detection, such as mackerel and milkfish, using the MobilenetV2. The results were able to produce accuracy of 97%, 94%, and 93% for each fish. The freshness detection method in this study has been implemented in the Fishku mobile-based application
Comparison of CNN Models With Transfer Learning in the Classification of Insect Pests
Full text linkInsect pests are an important problem to overcome in agriculture. The purpose of this research is to classify insect pests with the IP-102 dataset using several CNN pre-trained models and choose which model is best for classifying insect pest data. The method used is the transfer learning method with a fine-tuning approach. Transfer learning was chosen because this technique can use the features and weights that have been obtained during the previous training process. Thus, computation time can be reduced and accuracy can be increased. The models used include Xception, MobileNetV3L, MobileNetV2, DenseNet-201, and InceptionV3. Fine-tuning and freeze layer techniques are also used to improve the quality of the resulting model, making it more accurate and better suited to the problem at hand. This study uses 75,222 image data with 102 classes. The results of this study are the DenseNet-201 model with fine-tuning produces an accuracy value of 70%, MobileNetV2 66%, MobileNetV3L 68%, InceptionV3 67%, Xception 69%. The conclusion of this study is that the transfer learning method with the fine-tuning approach produces the highest accuracy value of 70% in the DenseNet-201 model
Analyze the Clustering and Predicting Results of Palm Oil Production in Aceh Utara
Full text linkPT. Perkebunan Nusantara 1 is engaged in oil palm production with a total land area of 1,144 Ha. The formulation of this research can determine productive land clusters based on land area, number of trees, number of stages, and palm oil production. Methodological steps include plantation area data and oil palm production data. This study can compare the C-means and K-means groups. As for predictions using the Backpropagation Neural Network (BPNN) algorithm and Fuzzy time series for production results. The results of grouping Cot girek palm oil production data for the 2019-2022 period from January to December were 1,365,530, while in 2022 it reached 1,768,720. The analysis used a land grouping method of 1,144 hectares, which resulted in 800.4 hectares of productive land and 343.6 hectares of less effective land. The results of the C-menas clustering model are more than K-meas with shorter iterations while for predictions it has an accuracy rate of 90.77%. As a comparison, the level of accuracy of the fuzzy time series is 81.27%. The results of this study can be used as recommendations for companies in the analysis of productive land grouping analysis and forecast results from these lands
Implementation of Ensemble Methods on Classification of CDK2 Inhibitor as Anti-Cancer Agent
Full text linkCancer is known as the second leading cause of death worldwide. About 7-10 million cases of death by cancer occur every year. The recent treatment to heal the cancer is chemotherapy. However, chemotherapy treatment is known to have side effects and cell resistance issues to certain drugs. Therefore, it is required to develop a new drug that can reduce the side effects and provide a better treatment effect. In general, anti-cancer drugs are developed by targeting Cyclin-Dependent Kinase 2 (CDK2) enzyme. Conventional drug design is not effective and efficient for obtaining new drug candidates because of no information about the biological activity before it is synthesized. In this study, we aim to develop a model to predict the activity of CDK2 inhibitors by using ensemble methods, i.e., XGBoost, Random Forest, and AdaBoost. The study was conducted by calculating several fingerprints, i.e., Estate, Extended, Maccs, and Pubchem, as feature variables. Based on the results, we found that Random Forest with Pubchem fingerprint gives the best result with the value of Matthews Correlation Coefficient (MCC) and Area Under the ROC Curve (AUC) values are 0.979 and 0.999, respectively. From this study, we contributed to revealing the potency of the ensemble with fingerprint in bioactivity prediction, especially CDK2 inhibitors as anti-cancer agents
Unsupervised Text Style Transfer for Authorship Obfuscation in Bahasa Indonesia
Full text linkAuthorship attribution is an NLP task to identify the author of a text based on stylometric analysis. On the other hand, authorship obfuscation aims to protect against authorship attribution by modifying a text’s style. The main challenge in authorship obfuscation is how to keep the content of the text despite the text modification. In this research, we are applying text style transfer methods for modifying the writing style while preserving the content of the input text. We implemented two unsupervised text style transfer: dictionary-based and back translation methods to change the formality level of the text. Experiment results shows that the back-translation method outperformed the dictionary-based method. The authorship attribution performance decreased up to 16.15% and 23.66% on F1-score for 3 and 10 authors respectively using back-translation. While for dictionary-based method the F1-score dropped up to 1.99% and 11.56% for 3 and 10 authors respectively. Evaluation on sensibleness and soundness factors show that the back-translation method can preserve the semantic of the obfuscated texts. Moreover, the modified texts are well-formed and inconspicuous.
Evaluation of Food Security Area of East Java Province Using Fuzzy C-Means (FCM) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS)
No full textThe formation of quality human resources cannot be separated from food, as nutritional intake affects human performance and health. As time increases, the number of residents increases to increase food needs. The ability of a region to meet food needs in its territory is different from other regions. This study aims to classify regions in East Java Province based on food security and determine areas with the best and lowest food security. The method used is the Fuzzy C-Means (FCM) and TOPSIS methods.This research uses criteria based on the Food Security Index compiled by the Food Security Agency. The results of regional clustering using FCM selected the best cluster using three clusters for all requirements, except in food utilization in the city using five clusters. Furthermore, from the clustering results, clustering and cluster members use TOPSIS and produce Magetan regency and Madiun city as areas with the highest food security. At the same time, the lowest food securities are Probolinggo regency and Kediri city
ESSAY ANSWER CLASSIFICATION WITH SMOTE RANDOM FOREST AND ADABOOST IN AUTOMATED ESSAY SCORING
Full text link Automated essay scoring (AES) is used to evaluate and assessment student essays are written based on the questions given. However, there are difficulties in conducting automatic assessments carried out by the system, these difficulties occur due to typing errors (typos), the use of regional languages , or incorrect punctuation. These errors make the assessment less consistent and accurate. Based on the dataset analysis that has been carried out, there is an imbalance between the number of right and wrong answers, so a technique is needed to overcome the data imbalance. Based on the literature, to overcome these problems, the Random Forest and AdaBoost classification algorithms can be used to improve the consistency of classification accuracy and the SMOTE method to overcome data imbalances.The Random Forest method using SMOTE can achieve an F1 measure of 99%, which means that the hybrid method can overcome the problem of imbalanced datasets that are limited to AES. The AdaBoost model with SMOTE produces the highest F1 measure reaching 99% of the entire dataset. The structure of the dataset is something that also affects the performance of the model. So the best model obtained in this study is the Random Forest model with SMOTE