Anggit Dwi Hartanto
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Anggit Dwi Hartanto has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Ema Utami
Universitas AMIKOM Yogyakarta
6 shared publications - Erwin Syahrudin
Universitas AMIKOM Yogyakarta
2 shared publications - Suwanto Raharjo
Institut Sains & Teknologi AKPRIND Yogyakarta
2 shared publications - Agung Budi Prasetio
Faculty Computer Science, Institut Teknologi Tangerang Selatan, Komplek Komersial BSD Kav 9, Jl. Raya Serpong, Lengkong Karya, Serpong Utara, Kota Tangerang Selatan 12246, Indonesia
1 shared publication - Ahmad Fauzi
Biology Education, Faculty of Teacher Training and Education, Universitas Muhammadiyah Malang, Indonesia
1 shared publication - Ahmad Fikri Iskandar
Program Studi Teknik Informatika, Univeristas Amikom Yogyakarta
1 shared publication
Latest Publications
- Open Access
- 12 July 2025
- Published by Universitas Nusantara PGRI Kediri in INTENSIF: Jurnal Ilmiah Penelitian dan Penerapan Teknologi Sistem Informasi
- Vol. 9 (2) , 178-194
Evaluating YOLOv8-Based Distance Estimation: A Comparison of OpenCV and Coordinate Attention Weighting in Blind Navigation Systems
Background: Recent developments in assistive technologies for the visually impaired have increasingly utilized computer vision techniques for real-time distance estimation. However, challenges remain in balancing accuracy, latency, and robustness under dynamic environmental conditions. Objective: This study aimed to evaluate and compare the performance of OpenCV and Coordinate Attention Weighting (CAW) models for distance estimation in blind navigation systems, particularly focusing on their effectiveness in real-time scenarios. Methods: A quantitative experimental study was conducted using an image dataset labeled with actual distances. The baseline performances of OpenCV and CAW were measured and compared. Subsequently, targeted optimizations were applied to the OpenCV model, including adaptive image filtering, hyperparameter tuning, and integration of a Kalman filter. Results: Initial evaluation showed that CAW achieved a higher baseline accuracy of 88% compared to OpenCV. However, after optimizations, OpenCV’s accuracy improved by 15%, reaching approximately 85%. Additionally, the optimized OpenCV model demonstrated reduced latency, outperforming CAW in real-time detection speed. Under varying lighting and motion conditions, OpenCV also exhibited superior robustness compared to CAW. Conclusion: The findings suggest that with proper optimization, OpenCV can match or exceed CAW in key performance aspects, making it a viable and efficient alternative for real-time distance estimation in blind navigation systems. Future research should explore further model integration and hardware acceleration for deployment in wearable devices.
- 12 July 2025
- Published by Universitas Nusantara PGRI Kediri in INTENSIF: Jurnal Ilmiah Penelitian dan Penerapan Teknologi Sistem Informasi
- Vol. 9 (2) , 178-194
Open AccessEvaluating YOLOv8-Based Distance Estimation: A Comparison of OpenCV and Coordinate Attention Weighting in Blind Navigation Systems
Background: Recent developments in assistive technologies for the visually impaired have increasingly utilized computer vision techniques for real-time distance estimation. However, challenges remain in balancing accuracy, latency, and robustness under dynamic environmental conditions. Objective: This study aimed to evaluate and compare the performance of OpenCV and Coordinate Attention Weighting (CAW) models for distance estimation in blind navigation systems, particularly focusing on their effectiveness in real-time scenarios. Methods: A quantitative experimental study was conducted using an image dataset labeled with actual distances. The baseline performances of OpenCV and CAW were measured and compared. Subsequently, targeted optimizations were applied to the OpenCV model, including adaptive image filtering, hyperparameter tuning, and integration of a Kalman filter. Results: Initial evaluation showed that CAW achieved a higher baseline accuracy of 88% compared to OpenCV. However, after optimizations, OpenCV’s accuracy improved by 15%, reaching approximately 85%. Additionally, the optimized OpenCV model demonstrated reduced latency, outperforming CAW in real-time detection speed. Under varying lighting and motion conditions, OpenCV also exhibited superior robustness compared to CAW. Conclusion: The findings suggest that with proper optimization, OpenCV can match or exceed CAW in key performance aspects, making it a viable and efficient alternative for real-time distance estimation in blind navigation systems. Future research should explore further model integration and hardware acceleration for deployment in wearable devices.
- Open Access
- 1 June 2025
- Published by Institute of Advanced Engineering and Science in International Journal of Electrical and Computer Engineering (IJECE)
- Vol. 15 (3) , 3267-3278
Comparative analysis of YOLOv8 techniques: OpenCV and coordinate attention weighting for distance perception in blind navigation systems
Blindness is a very important issue to consider in research aimed at assisting vision. This condition requires further study to provide solutions for the blind. This study evaluates and compares the effectiveness of the you only look once v8 (YOLOv8) model integrated with OpenCV and the coordinate attention weighting (CAW) technique for distance estimation in a blind navigation system. Initially, YOLOv8 integrated with OpenCV produced less than optimal results, prompting further improvement efforts to surpass the performance of CAW. The goal is to enhance the accuracy and efficiency of distance perception without the need for additional sensors. The materials used include a variety of datasets annotated with distance information to train and evaluate the model. The methods employed include integrating YOLOv8 with OpenCV for baseline comparison and applying CAW to improve distance perception through enhanced feature attention. The results show that YOLOv8+OpenCV Improved achieves the lowest mean squared error (MSE) across the entire distance range: 0-1 m (0.44), 1-2 m (0.50), 2-3 m (0.58), 3-4 m (0.64), and 4-5 m (0.71). YOLOv8+CAW also outperforms YOLOv8+OpenCV original, demonstrating a notable enhancement in accuracy. The model achieves a detection accuracy of 95.7%, showcasing the effectiveness of computer vision techniques in supporting blind navigation systems, offering precise distance estimation capabilities and reducing the reliance on external sensors. The implications include improved real-time performance and accessibility for the blind, paving the way for more efficient and reliable navigation assistance technologies.
- 1 June 2025
- Published by Institute of Advanced Engineering and Science in International Journal of Electrical and Computer Engineering (IJECE)
- Vol. 15 (3) , 3267-3278
Open AccessComparative analysis of YOLOv8 techniques: OpenCV and coordinate attention weighting for distance perception in blind navigation systems
Blindness is a very important issue to consider in research aimed at assisting vision. This condition requires further study to provide solutions for the blind. This study evaluates and compares the effectiveness of the you only look once v8 (YOLOv8) model integrated with OpenCV and the coordinate attention weighting (CAW) technique for distance estimation in a blind navigation system. Initially, YOLOv8 integrated with OpenCV produced less than optimal results, prompting further improvement efforts to surpass the performance of CAW. The goal is to enhance the accuracy and efficiency of distance perception without the need for additional sensors. The materials used include a variety of datasets annotated with distance information to train and evaluate the model. The methods employed include integrating YOLOv8 with OpenCV for baseline comparison and applying CAW to improve distance perception through enhanced feature attention. The results show that YOLOv8+OpenCV Improved achieves the lowest mean squared error (MSE) across the entire distance range: 0-1 m (0.44), 1-2 m (0.50), 2-3 m (0.58), 3-4 m (0.64), and 4-5 m (0.71). YOLOv8+CAW also outperforms YOLOv8+OpenCV original, demonstrating a notable enhancement in accuracy. The model achieves a detection accuracy of 95.7%, showcasing the effectiveness of computer vision techniques in supporting blind navigation systems, offering precise distance estimation capabilities and reducing the reliance on external sensors. The implications include improved real-time performance and accessibility for the blind, paving the way for more efficient and reliable navigation assistance technologies.
- Open Access
- 8 April 2025
- Published by Universitas Hamzanwadi in EDUMATIC: Jurnal Pendidikan Informatika
- Vol. 9 (1) , 149-158
Analisis Perbandingan Algoritma SVM, Random Forest dan Logistic Regression untuk Prediksi Stunting Balita
The prevalence of stunting in Banjarmasin City in 2023 reached 26.5%, exceeding the WHO target (below 20%). Stunting impacts physical growth, cognitive development, and long-term economic productivity. The purpose of this study is to compare the performance of SVM, random forest, and logistic regression algorithms in classifying the stunting status of toddlers. The approach we use is comparative quantitative with machine learning methods for health data classification. Data totaling 2,231 under-five records were obtained from the Banjarmasin City Health Office. We used age, weight, height, and z-score information. Data preprocessing includes handling missing values, categorical data transformation, numerical data standardization, and feature selection. The dataset was divided into 70:30 and 80:20 ratios using stratified sampling with 5-fold cross-validation. Our results show that SVM is the best model, with accuracy 92%, precision 91%, recall 99%, F1-score 95%, and AUC 99%, followed by random forest (accuracy 91%, AUC 98%) and logistic regression (accuracy 92%, AUC 97%). SVM showed superior performance due to its ability to find the optimal hyperplane that maximally separates stunted and non-stunted classes, as well as its effectiveness in handling non-linear data through kernel tricks. SVM's good generalization ability on new data makes it a top choice as a predictive tool for stunting prevention in Banjarmasin City.
- 8 April 2025
- Published by Universitas Hamzanwadi in EDUMATIC: Jurnal Pendidikan Informatika
- Vol. 9 (1) , 149-158
Open AccessAnalisis Perbandingan Algoritma SVM, Random Forest dan Logistic Regression untuk Prediksi Stunting Balita
The prevalence of stunting in Banjarmasin City in 2023 reached 26.5%, exceeding the WHO target (below 20%). Stunting impacts physical growth, cognitive development, and long-term economic productivity. The purpose of this study is to compare the performance of SVM, random forest, and logistic regression algorithms in classifying the stunting status of toddlers. The approach we use is comparative quantitative with machine learning methods for health data classification. Data totaling 2,231 under-five records were obtained from the Banjarmasin City Health Office. We used age, weight, height, and z-score information. Data preprocessing includes handling missing values, categorical data transformation, numerical data standardization, and feature selection. The dataset was divided into 70:30 and 80:20 ratios using stratified sampling with 5-fold cross-validation. Our results show that SVM is the best model, with accuracy 92%, precision 91%, recall 99%, F1-score 95%, and AUC 99%, followed by random forest (accuracy 91%, AUC 98%) and logistic regression (accuracy 92%, AUC 97%). SVM showed superior performance due to its ability to find the optimal hyperplane that maximally separates stunted and non-stunted classes, as well as its effectiveness in handling non-linear data through kernel tricks. SVM's good generalization ability on new data makes it a top choice as a predictive tool for stunting prevention in Banjarmasin City.
- Open Access
- 30 January 2023
- Published by Universitas Pahlawan Tuanku Tambusai in Journal on Education
- Vol. 5 (3) , 7925-7937
DDoS Penerapan Random Forest dan Adaboost untuk Klasifikasi Serangan DDoS
Di antara berbagai jenis serangan di bidang Teknologi Informasi, serangan DDOS adalah salah satu ancamanterbesar bagi situs internet dan menimbulkan risiko yang menghancurkan keamanan sistem komputer, terutamakarena potensi dampaknya. Oleh karena itu mengapa penelitian di bidang ini berkembang pesat, dengan parapeneliti yang berfokus pada cara-cara baru untuk mengatasi deteksi dan pencegahan intrusi. Machine learningdan Artificial Intelligent adalah beberapa tambahan terbaru dalam daftar teknologi yang diteliti untukmelakukan klasifikasi deteksi intrusi. Studi ini mengeksplorasi perilaku dan penerapan dataset DDoS untukpembelajaran mesin dalam konteks deteksi intrusi. Alur dalam penelitian ini, pertama adalah mengumpulkandataset DDoS mentah dari sumber yang memiliki reputasi baik. Setelah data diperoleh, kumpulan data akhirdibuat untuk pemodelan. Manajemen data melibatkan pembersihan data, transformasi tipe data dan pertukarandata pada pengumpulan data. Proses seleksi disertai dengan model. Dua algoritma terpisah, random danadaboost, digunakan untuk melatih model dengan dataset. Model divalidasi dan dilatih ulang dengan k-foldcross. Model tersebut akhirnya dievaluasi menggunakan data yang tidak terlihat. Hasilnya ditentukan olehberbagai ukuran keluaran. Dalam percobaan, dataset DDoS digunakan: CICDDoS_2019 Performa deteksiintrusi set data ini dianalisis menggunakan dua model pembelajaran mesin. Dataset dibagi dalam rasio 80:20untuk pelatihan model, validasi dan pengujian. Model pembelajaran mesin dipilih secara sistematis dan hatihatiuntuk memastikan bahwa eksperimen dilakukan dengan cara yang tepat. Hasilnya dianalisis menggunakansekumpulan metrik performa, termasuk akurasi, presisi, recall, f-measure, dan waktu komputasi.
- 30 January 2023
- Published by Universitas Pahlawan Tuanku Tambusai in Journal on Education
- Vol. 5 (3) , 7925-7937
Open AccessDDoS Penerapan Random Forest dan Adaboost untuk Klasifikasi Serangan DDoS
Di antara berbagai jenis serangan di bidang Teknologi Informasi, serangan DDOS adalah salah satu ancamanterbesar bagi situs internet dan menimbulkan risiko yang menghancurkan keamanan sistem komputer, terutamakarena potensi dampaknya. Oleh karena itu mengapa penelitian di bidang ini berkembang pesat, dengan parapeneliti yang berfokus pada cara-cara baru untuk mengatasi deteksi dan pencegahan intrusi. Machine learningdan Artificial Intelligent adalah beberapa tambahan terbaru dalam daftar teknologi yang diteliti untukmelakukan klasifikasi deteksi intrusi. Studi ini mengeksplorasi perilaku dan penerapan dataset DDoS untukpembelajaran mesin dalam konteks deteksi intrusi. Alur dalam penelitian ini, pertama adalah mengumpulkandataset DDoS mentah dari sumber yang memiliki reputasi baik. Setelah data diperoleh, kumpulan data akhirdibuat untuk pemodelan. Manajemen data melibatkan pembersihan data, transformasi tipe data dan pertukarandata pada pengumpulan data. Proses seleksi disertai dengan model. Dua algoritma terpisah, random danadaboost, digunakan untuk melatih model dengan dataset. Model divalidasi dan dilatih ulang dengan k-foldcross. Model tersebut akhirnya dievaluasi menggunakan data yang tidak terlihat. Hasilnya ditentukan olehberbagai ukuran keluaran. Dalam percobaan, dataset DDoS digunakan: CICDDoS_2019 Performa deteksiintrusi set data ini dianalisis menggunakan dua model pembelajaran mesin. Dataset dibagi dalam rasio 80:20untuk pelatihan model, validasi dan pengujian. Model pembelajaran mesin dipilih secara sistematis dan hatihatiuntuk memastikan bahwa eksperimen dilakukan dengan cara yang tepat. Hasilnya dianalisis menggunakansekumpulan metrik performa, termasuk akurasi, presisi, recall, f-measure, dan waktu komputasi.
- Open Access
- 19 June 2022
- Published by Universitas Hamzanwadi in EDUMATIC: Jurnal Pendidikan Informatika
- Vol. 6 (1) , 143-149
Comparison of Naïve Bayes Algorithm and XGBoost on Local Product Review Text Classification
Online reviews are critical in supporting purchasing decisions because, with the development of e-commerce, there are more and more fake reviews, so more and more consumers are worried about being deceived in online shopping. Sentiment analysis can be applied to Marketplace product reviews. This study aims to compare the two categories of Naïve Bayes and XGBoost by using the two vector spaces wod2vec and TFIDF. The methods used in this research are data collection, data cleaning, data labelling, data pre-processing, classification and evaluation. The data scraping process produced 25,581 data which was divided into 80% training data and 20% test data. The data is divided into two classes, namely good sentiment and bad sentiment. Based on the research that has been done, the combination of Word2vec + XGBoost F1 scores higher by 0.941, followed by TF-IDF + XGBoost by 0.940. Meanwhile, Naïve Bayes has an F1-Score of 0.915 with TF-IDF and 0.900 with word2vec. Classification using XGBoost proved to be able to classify unbalanced data better than Naïve Bayes.
- 19 June 2022
- Published by Universitas Hamzanwadi in EDUMATIC: Jurnal Pendidikan Informatika
- Vol. 6 (1) , 143-149
Open AccessComparison of Naïve Bayes Algorithm and XGBoost on Local Product Review Text Classification
Online reviews are critical in supporting purchasing decisions because, with the development of e-commerce, there are more and more fake reviews, so more and more consumers are worried about being deceived in online shopping. Sentiment analysis can be applied to Marketplace product reviews. This study aims to compare the two categories of Naïve Bayes and XGBoost by using the two vector spaces wod2vec and TFIDF. The methods used in this research are data collection, data cleaning, data labelling, data pre-processing, classification and evaluation. The data scraping process produced 25,581 data which was divided into 80% training data and 20% test data. The data is divided into two classes, namely good sentiment and bad sentiment. Based on the research that has been done, the combination of Word2vec + XGBoost F1 scores higher by 0.941, followed by TF-IDF + XGBoost by 0.940. Meanwhile, Naïve Bayes has an F1-Score of 0.915 with TF-IDF and 0.900 with word2vec. Classification using XGBoost proved to be able to classify unbalanced data better than Naïve Bayes.
- Open Access
- 20 December 2021
- Published by Universitas Hamzanwadi in EDUMATIC: Jurnal Pendidikan Informatika
- Vol. 5 (2) , 167-176
Perbandingan Performansi Model pada Algoritma K-NN terhadap Klasifikasi Berita Fakta Hoaks Tentang Covid-19
During Covid-19 pandemic, there was various hoax news about Covid-19. There are truth-clarification platforms for hoax news about Covid-19 such as Jala Hoax and Saber Hoax which categorize into misinformation and disinformation. Classification of supervised learning methods is applied to carry out learning from fact labels. Dataset is taken from Jala Hoax and Saber Hoax as many as 559 data which are made into Class 1 (Misleading Content, Satire/Parody, False Connection), Class 2 (False Context, Imposter Content), Class 3 (Fabricated and Manipulated Content). K-Nearest Neighbor (K-NN) is used to classify categories of misinformation and disinformation. Dissimilarity measure Jaccard Distance is compared with Euclidean, Manhattan, and Minkowski and uses k-value variance in the K-NN to determine the performance comparison results for each test. Results of Jaccard Distance at the value of k = 4 get a higher value than other model with an accuracy 0.696, precision 0.710, recall 0.572, and F1-Score. Maximum results tend to be on label of the most data class in Class 1 (Misleading Content, Satire or Parody, False Connection) with a total of 58 correct data from 61 test data.
- 20 December 2021
- Published by Universitas Hamzanwadi in EDUMATIC: Jurnal Pendidikan Informatika
- Vol. 5 (2) , 167-176
Open AccessPerbandingan Performansi Model pada Algoritma K-NN terhadap Klasifikasi Berita Fakta Hoaks Tentang Covid-19
During Covid-19 pandemic, there was various hoax news about Covid-19. There are truth-clarification platforms for hoax news about Covid-19 such as Jala Hoax and Saber Hoax which categorize into misinformation and disinformation. Classification of supervised learning methods is applied to carry out learning from fact labels. Dataset is taken from Jala Hoax and Saber Hoax as many as 559 data which are made into Class 1 (Misleading Content, Satire/Parody, False Connection), Class 2 (False Context, Imposter Content), Class 3 (Fabricated and Manipulated Content). K-Nearest Neighbor (K-NN) is used to classify categories of misinformation and disinformation. Dissimilarity measure Jaccard Distance is compared with Euclidean, Manhattan, and Minkowski and uses k-value variance in the K-NN to determine the performance comparison results for each test. Results of Jaccard Distance at the value of k = 4 get a higher value than other model with an accuracy 0.696, precision 0.710, recall 0.572, and F1-Score. Maximum results tend to be on label of the most data class in Class 1 (Misleading Content, Satire or Parody, False Connection) with a total of 58 correct data from 61 test data.
- 1 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 868-871
The Effect of Data Acquisition Techniques in Profiling Analysis Based on Twitter
Twitter have a tremendous number of tweets from its user. The bigger of the data’s volume and more complete the tweet’s data that gathered, the more accurate the personality analysis will be. This study is sought to analyze and measure which acquisition technique that viable to be used on the personality profile analysis and to know whether the acquisition technique will always return a huge amount of data volume in a longer time period with a faster time or not. The most commonly used acquisition technique is crawling, streaming and scrapping. This study makes uses of those three acquisition methods to measure the compatibility toward the personality profile analysis. Based on the previous study, the usage of scrapping acquisition method is proved to be an arguably good method as it can acquire a more completed tweet data in a huge amount with bigger computation time and the acquisition time is faster. It caused by the big volume of the data and it need more computational time, so the acquisition speed of the tweet data is faster. Based on that result, scrapping technique is said to be a good option to analyze someone’s personality profile.
- 1 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 868-871
The Effect of Data Acquisition Techniques in Profiling Analysis Based on Twitter
Twitter have a tremendous number of tweets from its user. The bigger of the data’s volume and more complete the tweet’s data that gathered, the more accurate the personality analysis will be. This study is sought to analyze and measure which acquisition technique that viable to be used on the personality profile analysis and to know whether the acquisition technique will always return a huge amount of data volume in a longer time period with a faster time or not. The most commonly used acquisition technique is crawling, streaming and scrapping. This study makes uses of those three acquisition methods to measure the compatibility toward the personality profile analysis. Based on the previous study, the usage of scrapping acquisition method is proved to be an arguably good method as it can acquire a more completed tweet data in a huge amount with bigger computation time and the acquisition time is faster. It caused by the big volume of the data and it need more computational time, so the acquisition speed of the tweet data is faster. Based on that result, scrapping technique is said to be a good option to analyze someone’s personality profile.