Article Per Year (5 Year)
p-Index From 2021 - 2026
2.303
P-Index
This Author published in this journals
All Journal Jurnal Ilmu Dasar BERKALA SAINSTEK JURNAL FISIKA Saintifika UNEJ e-Proceeding Panrita Abdi - Jurnal Pengabdian pada Masyarakat CogITo Smart Journal Jurnal Pendidikan Fisika dan Keilmuan (JPFK) JOURNAL OF APPLIED INFORMATICS AND COMPUTING Data Science: Journal of Computing and Applied Informatics JURTEKSI JURNAL GEOGRAFI Geografi dan Pengajarannya Jurnal Pengabdian dan Kewirausahaan NEAR: Jurnal Pengabdian kepada Masyarakat SAINTIFIKA Jurnal Pengabdian Masyarakat Nusantara
Puguh Hiskiawan
Universitas Bunda Mulia
Agriculture, Biological Sciences & Forestry Arts Humanities Astronomy Biochemistry, Genetics & Molecular Biology Chemical Engineering, Chemistry & Bioengineering Chemistry Civil Engineering, Building, Construction & Architecture Computer Science & IT Control & Systems Engineering Decision Sciences, Operations Research & Management Economics, Econometrics & Finance Education Electrical & Electronics Engineering Environmental Science Health Professions Industrial & Manufacturing Engineering Languange, Linguistic, Communication & Media Library & Information Science Mathematics Medicine & Pharmacology Nursing Physics Public Health Social Sciences Transportation Other

Published : 30 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 30 Documents
Search

 1   2   3 
MULTI VIEW FEATURE FUSION FOR INDUSTRIAL ANOMALY DETECTION USING 1D-CNN Daniel Fernando Nainggolan; Puguh Hiskiawan
JURTEKSI (jurnal Teknologi dan Sistem Informasi) Vol. 12 No. 2 (2026): Maret 2026
Publisher : Lembaga Penelitian dan Pengabdian Kepada Masyarakat (LPPM) STMIK Royal Kisaran

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33330/jurteksi.v12i2.4501

Abstract

Abstract: Anomalous sound detection is essential for industrial predictive maintenance, as machine failures often originate from subtle acoustic changes during operation. However, high background noise and limitations of conventional Convolutional Neural Networks (CNN) reduce detection reliability. This study proposes a 1D-CNN-based anomaly detection framework with multi-view feature fusion and temporal segmentation to enhance detection performance. The approach combines MFCC, Log-Mel Spectrogram, and Chroma STFT features, while temporal segmentation divides audio signals into 5-second segments to better capture transient anomalies. Experiments on the MIMII dataset under varying Signal-to-Noise Ratio (SNR) conditions show that MFCC and Log-Mel fusion achieves the best performance, with 97.90% accuracy and ROC-AUC of 0.9789. The model maintains accuracy above 90% at −6 dB, demonstrating strong robustness in noisy industrial environments. Keywords: industrial anomaly detection; 1D-CNN; multi-view feature fusion; temporal segmentation; MIMII dataset. Abstrak: Deteksi anomali suara merupakan komponen penting dalam sistem pemeliharaan prediktif industri, karena kegagalan mesin sering diawali oleh perubahan akustik yang bersifat halus selama proses operasi. Namun, tingkat kebisingan yang tinggi serta keterbatasan arsitektur Convolutional Neural Network (CNN) konvensional dapat menurunkan keandalan deteksi. Penelitian ini bertujuan mengusulkan kerangka deteksi anomali berbasis 1D-CNN yang mengintegrasikan strategi fusi fitur multi-view dan segmentasi temporal untuk meningkatkan kinerja deteksi. Pendekatan yang digunakan menggabungkan fitur MFCC, Log-Mel Spectrogram dan Chroma STFT, sementara teknik temporal splitting membagi sinyal audio menjadi segmen berdurasi 5 detik untuk menangkap anomali yang bersifat sementara. Eksperimen menggunakan dataset MIMII pada berbagai kondisi Signal-to-Noise Ratio (SNR) menunjukkan bahwa kombinasi MFCC dan Log-Mel Spectrogram menghasilkan kinerja terbaik dengan akurasi 97,90% dan ROC-AUC sebesar 0,9789. Model juga mempertahankan akurasi di atas 90% pada kondisi kebisingan ekstrem (−6 dB) yang menunjukkan ketahanan yang baik dalam lingkungan industri yang bising. Kata kunci: deteksi anomali industri; 1D-CNN; fusi fitur multi-view; segmentasi temporal; dataset MIMII
Applied Data Science Framework for Incremental and Interpretable Childhood Growth Risk Screening Puguh Hiskiawan; Theresia Puspa Wijayanti; Srava Chrisdes Antoro; Metta Gautama
Data Science: Journal of Computing and Applied Informatics Vol. 10 No. 2 (2026): Data Science: Journal of Computing and Applied Informatics (JoCAI) In Press
Publisher : Talenta Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32734/jocai.v10i2.24956

Abstract

Childhood growth monitoring plays an important role in the early identification of children who may experience developmental risks related to nutrition and health conditions. Conventional screening methods typically rely on anthropometric measurements that may not always be consistently obtained in community-based health environments. This study proposes an applied data science framework for incremental and interpretable screening of childhood growth risk using pose-derived body structure features combined with demographic and anthropometric attributes. Pose landmarks are extracted using the MediaPipe framework and integrated with variables including age, gender, and body weight to construct predictive models. Several machine learning algorithms are evaluated, including Logistic Regression, Random Forest, Gradient Boosting, XGBoost, K-Nearest Neighbors, and a Soft Voting Ensemble. Experimental evaluation using five-fold stratified cross-validation demonstrates that Logistic Regression achieves the highest predictive performance with a mean ROC-AUC of 0.901. Ablation analysis further indicates that incorporating pose-derived landmarks significantly improves classification performance compared with using demographic attributes alone. Interpretability analysis based on odds ratios highlights the contribution of pose features and demographic variables to prediction outcomes.
Benchmarking Pseudo-Mask Generation Methods for ResNet34-U-Net Acne Segmentation Nikita Amelia Valencia; Puguh Hiskiawan
Journal of Applied Informatics and Computing Vol. 10 No. 4 (2026): August 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i4.13568

Abstract

Acne lesion segmentation is essential for automated dermatological image analysis. However, developing accurate deep learning segmentation models requires pixel-level annotations, which are costly and time-consuming to obtain. To address this limitation, pseudo-mask generation methods can be utilized to produce surrogate labels for training segmentation networks without manual annotation. This study presents a comparative evaluation of five pseudo-mask generation methods for acne lesion segmentation, namely Contour-Based, Superpixel (SLIC), K-Means, Weakly Supervised Semantic Segmentation (WSSS) based on Otsu Thresholding and Morphological Operations, and Pseudo-Mask-Based Generation. The generated pseudo masks were employed as supervisory labels to train a ResNet34-U-Net segmentation model under identical experimental settings. To improve the robustness of the training process, data augmentation was applied exclusively to the training dataset. Segmentation performance was quantitatively evaluated using Dice Score, Intersection over Union (IoU), Precision, Recall and Validation Loss, and qualitatively assessed through visual comparisons of the generated pseudo masks and predicted segmentation results. The experimental results demonstrate that pseudo-mask quality has a substantial impact on segmentation performance. Among the evaluated methods, Superpixel (SLIC) achieved the highest performance with a Dice Score of 0.898, an IoU of 0.815, a Precision of 0.889, a Recall of 0.908, and the lowest validation loss of 0.268 indicating superior lesion boundary preservation and region consistency. WSSS (Otsu + Morphology) also produced competitive results, whereas Pseudo-Mask-Based and Contour-Based methods yielded comparatively lower performance. These findings demonstrate that high-quality pseudo masks can provide effective supervision for ResNet34-U-Net training, offering an annotation-efficient approach for acne lesion segmentation and providing practical insights into selecting suitable pseudo-mask generation methods for dermatological image analysis.
Integrasi AI dalam Dunia Kreatif melalui Pelatihan Sentiment Analysis untuk Penguatan Kompetensi Digital Siswa SMK DKV Atisa Dipamkara Puguh Hiskiawan; Jean Christopher Allen; Virananda Iskandar; Kunsuswandono Suryo Bagus Sutojo
Panrita Abdi - Jurnal Pengabdian pada Masyarakat Vol. 10 No. 3 (2026): Jurnal Panrita Abdi - Juli 2026
Publisher : LP2M Universitas Hasanuddin

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20956/pa.v10i3.48765

Abstract

This Community Service Program aimed to strengthen the digital competencies of Visual Communication Design students at Atisa Dipamkara Vocational High School through training and interactive practicum on Sentiment Analysis based on Artificial Intelligence (AI) within the field of Natural Language Processing (NLP). The program was implemented using a participatory and hands-on learning approach that combined short lectures, technical demonstrations, guided practice, and assisted practicum activities. Case studies were drawn from movie reviews, online product comments, and short social media texts to familiarize students with authentic digital content commonly encountered in the creative industry. A total of 40 students participated in the program from start to completion. Learning outcomes were evaluated through pre-test and post-test instruments based on four competency dimensions: educative (knowledge), objective (concept mastery), accountable (technical skills), and transparent (ability to interpret and communicate analytical results). The evaluation results showed that all competency dimensions experienced score improvements exceeding the minimum success indicator, with the transparent dimension recording the highest increase (33%), followed by educative (26%), objective (25%), and accountable (16%). These findings indicate that the training and interactive practicum model effectively enhanced students’ conceptual understanding of AI technologies, analytical thinking skills, confidence in using digital tools, and readiness to engage in a data-driven creative industry ecosystem. Therefore, this program can be considered a replicable and sustainable community service model for integrating AI-based learning into creativity-oriented vocational education.   Kegiatan Pengabdian kepada Masyarakat ini bertujuan untuk memperkuat kompetensi digital siswa Program Studi Desain Komunikasi Visual (DKV) SMK Atisa Dipamkara melalui pelatihan dan praktikum interaktif Sentiment Analysis berbasis Artificial Intelligence (AI) pada ranah Natural Language Processing (NLP). Kegiatan dilaksanakan dengan pendekatan pembelajaran partisipatif dan aplikatif yang memadukan pemaparan konsep dasar, demonstrasi teknis, latihan terbimbing, serta pendampingan praktikum. Studi kasus yang digunakan meliputi ulasan film, komentar produk daring, dan teks singkat media sosial untuk membiasakan siswa menganalisis teks digital yang relevan dengan konteks industri kreatif. Sebanyak 40 siswa mengikuti seluruh rangkaian kegiatan dari awal hingga akhir. Capaian pembelajaran dievaluasi melalui pre-test dan post-test berdasarkan empat dimensi kompetensi, yaitu edukatif (pengetahuan), objektif (penguasaan konsep), akuntabel (keterampilan teknis), dan transparan (kemampuan menginterpretasikan serta menyampaikan hasil analisis). Hasil evaluasi menunjukkan bahwa seluruh dimensi kompetensi mengalami peningkatan skor yang melampaui indikator keberhasilan yang telah ditetapkan, dengan dimensi transparan mencatat peningkatan tertinggi sebesar 33%, diikuti oleh edukatif sebesar 26%, objektif 25%, dan akuntabel 16%. Temuan ini menunjukkan bahwa model pelatihan dan praktikum interaktif berbasis NLP efektif dalam meningkatkan literasi AI, kemampuan berpikir analitis, serta kesiapan siswa vokasi untuk berpartisipasi dalam ekosistem industri kreatif berbasis data. Dengan demikian, program ini berpotensi menjadi model pengabdian berkelanjutan dan replikatif dalam penerapan AI pada pendidikan vokasi berbasis kreativitas
An Applied Data Science Approach for Detecting Depression Symptoms in Indonesian Social Media Text Using Transformer Models Winson, Winson; Hiskiawan, Puguh
Journal of Applied Informatics and Computing Vol. 10 No. 3 (2026): June 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i3.12644

Abstract

Depression is a mental health disorder that often remains undetected due to limited access to mental health services and persistent social stigma. Social media platforms provide an alternative source for identifying depressive symptoms through linguistic expressions shared by users in textual posts. This study proposes an applied data science approach for detecting depression symptoms in Indonesian social media text using Transformer-based models. The dataset was constructed by combining the DEPTWEET dataset with social media posts collected through keyword-based scraping guided by PHQ-9 indicators. The proposed framework consists of dataset construction, text preprocessing, Transformer-based modeling, and performance evaluation. Two pre-trained language models, IndoBERT and XLM-RoBERTa, were evaluated under two preprocessing configurations, namely normal preprocessing and light preprocessing. Experimental results show that preprocessing strategies significantly influence classification performance. Light preprocessing consistently improves contextual representation and leads to better results compared with normal preprocessing. XLM-RoBERTa combined with light preprocessing achieves the best overall performance with a test accuracy of 0.77 and an F1-score of 0.77. Additional robustness analysis and pairwise model agreement evaluation further indicate that both models maintain relatively stable predictions when processing noisy social media text. Findings from this study demonstrate the effectiveness of Transformer-based models for multi-class depression detection in Indonesian social media environments. The proposed framework provides insights into how applied data science techniques can support large-scale analysis of mental health signals in online platforms and contribute to the development of data-driven approaches for early detection of depression symptoms.
Performance Evaluation of YOLO26 for Pothole Detection Under Diverse Environmental Conditions Stephanie, Edelin; Hiskiawan, Puguh
Journal of Applied Informatics and Computing Vol. 10 No. 5 (2026): October 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i5.13950

Abstract

Road potholes are a serious threat to road safety. They can damage vehicles and increase the risk of traffic accidents. However, factors like lighting and weather can impact how well computer vision systems detect potholes. To tackle this issue, this study compares four You Only Look Once (YOLO) 2026 (YOLO26) variants: YOLO26n, YOLO26s, YOLO26m, and YOLO26l, for detecting potholes in different environmental conditions. A dataset was independently collected in the Poris area of Tangerang, Indonesia, using a forward-facing camera to capture a driver's view. The video recordings were turned into annotated image frames in YOLO format, creating 368 images that included clear, rain, evening, and night conditions. Each model was trained for 200 epochs under the same experimental conditions. Their performance was evaluated using Precision, Recall, mean Average Precision at an Intersection over Union (IoU) threshold of 0.5 (mAP@0.5), and mean Average Precision across IoU thresholds from 0.5 to 0.95 (mAP@0.5:0.95). Computational efficiency was evaluated using parameter count, model size, training time, GPU memory usage, inference time, and frames per second (FPS) to assess the computational requirements and real-time detection capability of each YOLO26 variant. The results showed that YOLO26m achieving the highest Recall of 0.877 and mAP@0.5 of 0.934, while YOLO26l achieved the highest Precision of 0.989 and mAP@0.5:0.95 of 0.735. On the other hand, YOLO26n was the most computationally efficient option while still providing competitive detection performance. These results highlight the trade-off between detection accuracy and computational efficiency among YOLO26 variants. They also offer useful guidance for choosing the right model for real-time pothole detection in various conditions, which can help in the development of future driver assistance systems and pothole warning systems.
CNN-LSTM with Scaled Dot-Product Attention for PM2.5 Forecasting in Jakarta: A Univariate Baseline Brilliant hendrylie; Puguh Hiskiawan
Journal of Applied Informatics and Computing Vol. 10 No. 5 (2026): October 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i5.13952

Abstract

This study evaluates a CNN-LSTM architecture with scaled dot-product attention for hour-ahead PM2.5 forecasting in Jabodetabek. Due to data constraints, the final model operates on a single univariate input channel - PM2.5's own first-order lag (t-1) - rather than the intended multivariate design; a supplementary experiment that added cyclical calendar attributes (hour, day-of-week, month) to the input did not improve accuracy and was therefore excluded from the final architecture. A three-member ensemble achieves R² = 0.788 and MAE = 9.654 µg/m³ on the held-out test set, outperforming a naive persistence baseline on R² (0.678) and RMSE, though the two are comparable on MAE, indicating the model's advantage lies chiefly in curbing large errors rather than typical-case error. Performance is strong in the dominant high-concentration regime but degrades sharply in the sparsely represented low-to-moderate range, and residual diagnostics indicate the model has not fully captured the series' temporal dependence. These results position the architecture as a working univariate baseline rather than an operationally validated forecaster, and motivate future work incorporating co-pollutant and meteorological data with per-city sequence construction to realize its full multivariate design.
Extraction of Transaction Information on Handwritten Notes Using Optical Character Recognition (OCR) and Named Entity Recognition (NER) Meyry Ardianti; Puguh Hiskiawan
Journal of Applied Informatics and Computing Vol. 10 No. 5 (2026): October 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i5.13985

Abstract

Handwritten notes remain widely used as transaction proofs, yet extracting information from them automatically is challenging due to variations in handwriting styles, layouts, and image qualities. This research aims to develop a transaction information extraction method for handwritten notes by integrating Optical Character Recognition (OCR) and Named Entity Recognition (NER). The research stages include image preprocessing using grayscale conversion, denoising, and Contrast Limited Adaptive Histogram Equalization (CLAHE), text recognition using EasyOCR across multiple image orientations, and transaction entity identification using IndoBERT with the BIO scheme. OCR performance is evaluated using Character Error Rate (CER) and Word Error Rate (WER), while NER is evaluated using Accuracy, Precision, Recall, and F1-score. Test results show that the NER model achieved an Accuracy of 0.8889, Precision of 0.7857, Recall of 0.8462, and F1-score of 0.8148 on the test data. However, the OCR-NER integration still exhibits character recognition errors and fails to completely extract all transaction attributes, particularly PRICE and AMOUNT. These findings indicate that the OCR-NER integration holds potential for supporting transaction data digitization in handwritten notes, though improving OCR quality and using a larger dataset are necessary to enhance system accuracy and generalizability.
Regime-Conditioned Causal Attribution and Forecasting of the Rupiah Jason Jesse Joel Polii; Puguh Hiskiawan
Journal of Applied Informatics and Computing Vol. 10 No. 5 (2026): October 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i5.14069

Abstract

This study asks whether the attribution of a predictive model of the Rupiah (IDR/USD) agrees with an independent structural decomposition, separately within volatility regimes. Hybrid econometric and machine-learning studies are evaluated almost exclusively by forecast error, leaving untested whether the learning stage weights the drivers the structural stage identifies. The pipeline combines a Structural Vector Autoregression (SVAR) with Cholesky identification, four-state volatility Markov-switching labeling, regime-conditioned features, walk-forward XGBoost, and SHAP attribution reconciled per regime with the Forecast Error Variance Decomposition (FEVD), on daily data from January 2010 to June 2026. The FEVD shows own-shock dominance of 92.5 percent at the 22-day horizon, robust to alternative orderings and to identification through heteroskedasticity (82.0 percent), though sub-period estimates range from 78.8 to 95.3 percent. On direction, XGBoost attains 56.4 percent at t+1 (Pesaran-Timmermann, p less than 0.001); on magnitude it does not beat a random walk at any horizon, whereas a GRU baseline does, so the predictive claim is confined to direction and gradient boosting is retained for its exact tree attribution. Exact permutation tests over all relabelings of the driver axis find agreement significant in only one of twelve regime-horizon cells, the tail regime at 22 days (cosine 0.902, p = 0.036), carried by 2020, while the own-dynamics share rises monotonically from 18.1 to 42.4 percent across regimes, significant at every horizon. The contribution is a falsifiable per-regime attribution audit with an explicit account of where it fails.
Visual Genealogy of Typography: Computational Provenance-Based Analysis of Typeface Evolution Hendi Thamrin; Shierly Everlin; Puguh Hiskiawan; Valentina Valentina; Angelyn Stevanie Lee; Paskalis Peter
Journal of Applied Informatics and Computing Vol. 10 No. 5 (2026): October 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i5.14211

Abstract

Computational analyses of typographic corpora increasingly produce similarity networks that are read as evidence of stylistic descent, yet no study has tested what such morphology-based genealogical inference can legitimately support; this study provides that test. Using a quantitative re-analysis design, we took an existing dataset of 33 typeface families, with 1,188 glyph images and eleven morphological descriptors, and added an auditable provenance layer recording designer, foundry, dates, revival status and script category. Five diagnostics were applied at family level: chronological correspondence, script-category composition tested with Fisher's exact test and the adjusted Rand index, multiple-comparison correction, classification stability, and a comparison of documented derivations with undocumented pairs. Eleven of sixteen families designated historical carry design dates after the declared period of 1450–1899, and no rendered file belongs to it. The era groups differ in serif composition before any morphology is measured (50.0% versus 5.9%, p = 0.0066); recovered clusters align with script category rather than period (adjusted Rand index 0.534 versus 0.090) under all corpus configurations examined; and among sans-serif families no era difference is detected on the descriptors testable (all p > 0.70). Only two of eleven descriptor differences survive family-wise correction, a headline accuracy of 75.8% falls to a mean near 0.68 under repeated validation against a 51.5% baseline, and undocumented pairs reach similarities above documented derivations. Morphological similarity therefore measures formal affinity conditioned on corpus composition, a necessary but not sufficient basis for claims of descent. A five-point reporting protocol for computational typography is proposed.
Co-Authors A.A. Ketut Agung Cahyawan W Adeline Yolanda C. Cristie Agustian Sukarno, Ferry Agustinus Fritz Wijaya Agustinus, Khemal Angelyn Stevanie Lee Artoto Arkundato Brilliant hendrylie Budidarma, Ni Gusti A. A. Evangelina Chandra, Earlene Aprilie Christina, Livia Daniel Fernando Nainggolan Evasaria Magdalena Sipayung Everlin, Shierly Fitri Rizqi Azizah Geasela, Yemima Geasela, Yemima Monica Hendi Thamrin Himmah Khasanah Illavi Praseti Pebrian Jason Jesse Joel Polii Jean Christopher Allen Kezia Josephine Kristina, Jatim Kunsuswandono Suryo Bagus Sutojo Merryana Lestari Metta Gautama Meyry Ardianti Nikita Amelia Valencia Nurul Piyantari Paskalis Peter Puspita, Maria Entina Rano Rano, Rano Reni Agustiani, Reni Ruliyanti Ruliyanti Srava Chrisdes Antoro Sri Hartanto Stephanie, Edelin Supriyadi Supriyadi Supriyadi Supriyadi Teguh Santoso Theresia Puspa Wijayanti Tio Jansel, Louis Feliepe Tipani Tio, Cynthia Valentina Valentina Vicky, Vicky Virananda Iskandar William, Jovan Winson, Winson Yemima Monica Geasela Yudhistira Adi Nugraha