2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)最新文献

{"title":"CRISP-DM for Data Quality Improvement to Support Machine Learning of Stunting Prediction in Infants and Toddlers","authors":"Ayi Purbasari, F. Rinawan, Arief Zulianto, A. Susanti, Hendra Komara","doi":"10.1109/ICAICTA53211.2021.9640294","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640294","url":null,"abstract":"Many Machine Learning (ML) projects ended up only as proof concept and failed to be produced. Therefore, this research focused on well-defined processes that must be followed, adopted the Cross-Industry Standard Process for Data Mining (CRISP-DM) with the specifications and requirements of supervised and unsupervised learning which includes a methodology for Classification/grouping. The Data Understanding and Data Preparation phases, used transactional data on examination of infants and toddlers in 2018-2021 on the iPosyandu application. At the Business Understanding stage, the ML was intended to predict stunting, so that data quality of iPosyandu can be informed and then recommendations and feature improvements and assistance for end-users can be made. The output of Data Understanding and Data Preparation was in the form of baby & toddler examination dataset, which was used in the Machine Learning modeling stage, especially to classify and predict nutritional/stunting status. Of the 192 tables contained in the iPosyandu application, there were 5 main tables that were needed to define the dataset. 75,652 data on infants and toddlers were checked with 49,615 data of examinations in 3173 Posyandu, which resulted in clean data of 39,411 rows of datasets for all examinations and 13,868 rows of datasets for the last examination of infants and toddlers. The dataset was combined with the nutritional status of infants and toddlers resulting from the calculation of the baby’s weight, length of the baby’s body, and the comparison of the baby’s height and weight. The dataset was tested into the ML using the Orange Application and produced a Classification model that can be used for prediction. From the results of the modeling evaluation, it can be seen that the Naïve Bayes Algorithm had an advantage with a predictive value of 0.851 while the Tree algorithm was 0.848 and the Neural Net was 0.845. From the overall evaluation, it can be concluded that there is a need to improve data quality by improving the application and improving the literacy of the end-users, so that the data has better quality and ready to be used as a ML dataset. The selected features can be aggregated to simplify the modeling process so as to obtain the expected model.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"181 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122633443","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Knowledge Graph Construction using Information Extraction of Indonesia Cosmetic Product Text in Bahasa Indonesia","authors":"Deborah Aprilia Josephine, A. Purwarianti, F. Ekaputra","doi":"10.1109/ICAICTA53211.2021.9640251","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640251","url":null,"abstract":"Knowledge graphs can be used for entity recognition in text, graph visualization, and to improve business processes, e.g. information retrieval in E-commerce. One of the information sources for building knowledge graphs is text data available in many digital system, such as E-commerce platform. In this paper, we proposed an approach to extract knowledge graph entities from product text available from E-commerce platforms. We utilize transfer learning technique with full-fine tuning from an existing trained model in order to recognize the entities due to the limitation of labeled data. Since some English terms are expressed in product texts, we used multilingual pretrained models with the Transformer Architecture, i.e. multilingual-BERT-base-cased (mBERT) and XLM-RoBERTa-base (XLMR) in our approach. The extracted entities were then mapped into a knowledge graph by adopting Text to Knowledge Graph (T2KG) framework components, i.e. using entity mapping and triple integration. The training data contains 1.500 labeled texts, while the test data contains 216 labeled texts conducted in three versions of data and four scenarios. Our evaluation result showed that the XLMR model performed better than mBERT for entity extraction task with an average F1-score of 0,895. Furthermore, we manually evaluate the knowledge graph mapping and construction using 1.445 product texts from two E-commerce platforms, which resulted in 338 entities formed in the knowledge graph with mapping precision 0,94.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"235 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122711812","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"3D Mesh Generation by Introducing Extended Attentive Normalization","authors":"Yuta Fukatsu, Masaki Aono","doi":"10.1109/ICAICTA53211.2021.9640290","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640290","url":null,"abstract":"In recent years, research on conditional image generation using GANs of the type where conditions are given by class labels or texts has been successful. On the other hand, the generation of conditional 3D models consisting of 3D meshes is still in its infancy. In this research, we add global information based on Attentive Normalization to local information using CNN to improve 3D mesh generation. Specifically, we propose Conditional Attentive Normalization, which is an extension of Attentive Normalization and can add conditional information. Comparative experiments conditioned by class labels and texts have been carried out by using Caltech-UCSD Birds-200-201. It turns out that our proposed method outperforms the conventional methods.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127722398","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vehicle Speed Estimation Using YOLO, Kalman Filter, and Frame Sampling","authors":"Asif Hummam Rais, R. Munir","doi":"10.1109/ICAICTA53211.2021.9640272","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640272","url":null,"abstract":"Vehicle speed estimation based on video feed can be used to enforce road rules and give traffic insights without the need of physical interference. Common methods are background subtraction, motion detection, and/or convolutional neural network (CNN). The first two methods suffer from inability to differentiate classes and occlusion, whereas CNN suffer from computational complexity. A system based on You Only Look Once (YOLO) as detector and Kalman filter as tracker is proposed. TensorRT and frame sampling are used to further optimize the inference time. From experiment using BrnoCompSpeed dataset on i5 9600k and RTX 2060 Super environment, proposed system runs the entire process at 118 FPS with mean average error (MAE) of 0.96 km/h and [-3,2] error interval at 93.81%. Frame sampling can be used to further improve the FPS, with 1/5 sampling improves the speed by 50% to 177 FPS with only 0.11 km/h MAE tradeoff to 1.07 km/h.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116754456","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Comparative Study on Language Models for Task-Oriented Dialogue Systems","authors":"Vinsen Marselino Andreas, Genta Indra Winata, A. Purwarianti","doi":"10.1109/ICAICTA53211.2021.9640249","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640249","url":null,"abstract":"The recent development of language models has shown promising results by achieving state-of-the-art performance on various natural language tasks by fine-tuning pre-trained models. In task-oriented dialogue (ToD) systems, language models can be used for end-to-end training without relying on dialogue state tracking to track the dialogue history but allowing the language models to generate responses according to the context given as input. This paper conducts a comparative study to show the effectiveness and strength of using recent pre-trained models for fine-tuning, such as BART and T5, on end-to-end ToD systems. The experimental results show substantial performance improvements after language model fine-tuning. The models produce more fluent responses after adding knowledge to the context that guides the model to avoid hallucination and generate accurate entities in the generated responses. Furthermore, we found that BART and T5 outperform GPT-based models in BLEU and F1 scores and achieve state-of-the-art performance in a ToD system.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"106 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116853935","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cost-Sensitive Learning and Ensemble BERT for Identifying and Categorizing Offensive Language in Social Media","authors":"Fajar Muslim, A. Purwarianti, F. Z. Ruskanda","doi":"10.1109/ICAICTA53211.2021.9640280","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640280","url":null,"abstract":"Some people often abuse freedom of expression on social media to carry out offensive actions. So we need mechanisms to keep social media conducive. This research aims to identify and categorize offensive language on social media, which consists of three subtasks: Offensive language identification (subtask A), Automatic categorization of offense types (subtask B), and Offense target identification (subtask C). These three subtasks use the OLID dataset (Offensive Language Identification Dataset) [1]. The previous research which utilized fine-tuning BERT achieved competitive performance at the OLID dataset (Zampieri. et al. 2019). In this research, we improve fine-tuning BERT performance using cost-sensitive learning and ensemble technique. The evaluation results on test data beat state of the art on subtask B (F1-score 0.7776), second position on subtask A (F1-score 0.8207), and second position on subtask C (F1-score 0.6574).","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124489850","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep Learning Based Spatiotemporal Human Action Recognition and Localization System","authors":"Jesslyn Nathania, N. P. Utama","doi":"10.1109/ICAICTA53211.2021.9640274","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640274","url":null,"abstract":"Spatiotemporal human action localization system is a field in computer vision and is of interest for real-world applications implemented in smart surveillance cameras, such as to improve public security, monitor patients' activities, or even detect any early symptoms of certain diseases. The system presented in this thesis followed the YOWO machine learning architecture reference, which was proposed by Köpüklü etc. (2019). YOWO extracts both spatial and temporal information. Bounding box regression and action classification can be done end-to-end. This aims to generate output faster compare to other state-of-the-art approaches. The implementation of this system is trained and tested with J-HMDB and NTU RGB+D datasets. Using certain specifications of machine defined, the system is just able to process video at 0.75seconds per frame with an accepted accuracy value. However, the system succeeds in increasing the human action localization accuracy from the YOWO reference with an accuracy of 41.6% to 43.84%.The result of the experiments shows that the modified architecture is able to improve the accuracy of YOWO. However, it slows down the frame rate of video processing.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122573383","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development Of Video Based Action Recognition System For Item Taking From Shelf","authors":"Irfan H. Widyadhana, D. H. Widyantoro","doi":"10.1109/ICAICTA53211.2021.9640283","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640283","url":null,"abstract":"This paper explores the development of video-based action recognition system for taking/returning item from/to the shelf that is fast enough to be used in real-time. For action classification, transfer learning is used on the Inflated 3D model with the ResNet50v1 backbone. Meanwhile, for object recognition, transfer learning is used with the YOLOv4 model because it includes components to determine the position of objects in the image. The system also adds a detector component so that the two classification components do not run continuously. Evaluation of the system produces an accuracy of 80% for the action classification component and 70% for the object recognition component. The entire system is capable of running on real-time video at speeds of up to 25 FPS.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"119 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125222953","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sentiment Analysis of Face-to-face Learning during Covid-19 Pandemic using Twitter Data","authors":"Ghanim Kanugrahan, A. Wicaksono","doi":"10.1109/ICAICTA53211.2021.9640282","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640282","url":null,"abstract":"Covid-19 pandemic has massive impacts on the activity of human in the world, including in Indonesia. To reduce the transmission of the virus, Indonesian government issues a policy to restrict daily public activities, affecting key national sectors, such as education systems. All learning activities are switched from the conventional face-to-face mode to being remote via the use of the Internet. After the pandemic begins to subside, the government then plans to reopen all schools and to allow face-to-face learning. However, this decision has sparked controversy in the social media, including Twitter. This paper describes a methodology to perform sentiment analysis on a collection of tweets that are in connection with the restart of the face-to-face learning mode. In particular, our experiments using hand-crafted features based on the tweets demonstrate that data-driven models are useful for automatic sentiment orientation classification on Twitter data. The best model achieved in this study has 69,1% accuracy, 68.6% precision, 69.1% recall, and 67,8% F1-Score. This result is achieved by using unigram, Support Vector Machine, and tweet + number of words (count) feature combinations.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"132 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131481847","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Semantic Segmentation based on Extended MobileNet with FPN","authors":"Yuki Sugimoto, Masaki Aono","doi":"10.1109/ICAICTA53211.2021.9640256","DOIUrl":"https://doi.org/10.1109/ICAICTA53211.2021.9640256","url":null,"abstract":"Recently, the development of Semantic Segmentation models has been actively pursued. Among them, many models are divided into Backbone module (encoder) and Head module (decoder). Feature Pyramid Network (FPN) is one of the methods to improve the accuracy. This is a module that performs feature extraction at multiple resolutions. By adding the FPN between the Backbone and the Head, we can perform fine segmentation of objects of various sizes. However, when FPNs are added to the segmentation model, the amount of computation required for prediction becomes huge. To solve this problem, we propose Reshaped Feature Pyramid Network (RFPN), which is a lightweight version of FPN. Furthermore, we propose an architecture for Segmentation that adds RFPN between Backbone and Head. In this architecture, we use MobileNet, which is a lightweight and highly accurate CNN, for the Backbone. In addition, we adopt Pyramid Pooling Module, which performs average pooling with various sizes, for the Head. The experimental results using PASCAL VOC 2012 show that the proposed method turns out to be a good trade-off between accuracy and computational time in terms of mean IoU and FLOPs, respectively.","PeriodicalId":217463,"journal":{"name":"2021 8th International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134266791","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}