cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
-
Contact Email
-
Phone
-
Journal Mail Official
-
Editorial Address
-
Location
,
INDONESIA
JOURNAL OF APPLIED INFORMATICS AND COMPUTING
Published by Politeknik Negeri Batam
ISSN : -     EISSN : 25486861     DOI : 10.3087
Core Subject : Science,
Computer Science & IT
Journal of Applied Informatics and Computing (JAIC) Volume 2, Nomor 1, Juli 2018. Berisi tulisan yang diangkat dari hasil penelitian di bidang Teknologi Informatika dan Komputer Terapan dengan e-ISSN: 2548-9828. Terdapat 3 artikel yang telah ditelaah secara substansial oleh tim editorial dan reviewer.
Arjuna Subject : -
Articles 1,006 Documents
 94   95   96   97   98 
Real-Time BISINDO Gesture Detection Using YOLOv8 for a Web-Based Text-to-Speech Prototype Riska Dewi Yuliyanti; M. Rafi Muttaqin; Teguh Iman Hermanto
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.v10i4.13362

Abstract

Communication challenges persist between the deaf and hard-of-hearing community and the general public, largely due to limited awareness and understanding of Indonesian Sign Language (BISINDO) in the broader population. This study develops a real-time web-based system that identifies BISINDO gestures using the YOLOv8 object detection model and converts the recognized gesture sequence into spoken words via a text-to-speech function. The study is based on the CRISP-ML(Q) framework, which includes stages such as understanding the data, preparing the data, building models, assessing their performance, implementing them in real-world applications, and continuously tracking their effectiveness. A total of 1,550 images were independently collected using a laptop camera and categorized into 31 classes, including 30 BISINDO gesture classes and 1 class for negative samples. The dataset was split using a stratified method, allocating 80% for training, 10% for validation, and 10% for testing. The YOLOv8 model was trained on Google Colaboratory using a T4 GPU runtime. The evaluation results indicate that the model achieved a precision of 0.979, a recall of 0.976, an mAP50 of 0.993, and an mAP50-95 of 0.839, which demonstrates robust performance in detecting BISINDO gestures. The developed model was incorporated into a web prototype, with FastAPI serving as the backend and HTML, CSS, and JavaScript utilized for the frontend. The system can identify hand gestures using a webcam, show bounding boxes with corresponding labels, compile the detected gestures into simple text sequences, and produce speech from that information. Latency testing revealed an average response time of around 210 milliseconds when the model was deployed locally and approximately 670 milliseconds when deployed online via Hugging Face Spaces. The system still faces challenges in identifying similar-looking gestures and is influenced by factors such as lighting, hand placement, and the availability of hosting resources. These results indicate that YOLOv8s is effective for detecting the 30 static BISINDO gesture classes evaluated in this study within a controlled data collection setting, though further validation is required before the system can be considered suitable for broader real-world assistive communication use.
Improving Retrieval-Augmented Generation Grounding Using Document Hierarchy Chunk Graph Putri Cristin; Hilmil Pradana
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.v10i4.13368

Abstract

Retrieval-Augmented Generation (RAG) is widely used to enhance question-answering systems across various domains. However, while real-world source documents are inherently structured, conventional RAG approaches primarily rely on semantic similarity between isolated text chunks, which can overlook document hierarchy and limit retrieval effectiveness. To address this issue, this study introduces a document hierarchy-based Chunk Graph approach to improve retrieval grounding in RAG systems. The proposed framework preserves document hierarchy during chunking and models inter-chunk relationships using a weighted graph that combines structural proximity and semantic similarity. The approach was evaluated using the StructuredQA and CUAD benchmark datasets, with performance measured via Precision, Recall, and F1-Score. Experimental results demonstrate that the effectiveness of the Chunk Graph depends heavily on the source document format. On the highly structured StructuredQA dataset, the proposed method successfully connects fragmented information, improving the F1-Score from 47.62% to 50.23%. Conversely, on the CUAD dataset which consists of raw text with implicit hierarchy and no nested structure, the model becomes redundant and does not yield performance gains. These findings conclude that integrating structural-semantic relationships significantly improves context selection, specifically for documents with explicit hierarchical structures.
Improving Software Effort Estimation Through Feature Selection and Optimized SVR Rahmi Putri; Gusti Eka Yuliastuti; Citra Nurina Prabiantissa
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.v10i4.13378

Abstract

Accurate software development effort estimation is essential but often hindered by high-dimensional data and the inefficiencies of handling feature selection and parameter tuning as separate, sequential processes. This study proposes an integrated Whale Optimization Algorithm–Support Vector Regression (WOA-SVR) framework that simultaneously optimizes binary feature selection and continuous SVR hyperparameters (C, γ, ϵ) within a unified search process. Evaluated using the NASA93 dataset under a strict nested 10-fold cross-validation protocol to prevent information leakage, the proposed model's performance was comprehensively assessed using six metrics (MMRE, MdMRE, Pred (25), RMSE, MAE, MAPE), accompanied by mean and standard deviation reporting. A rigorous ablation study empirically proved that simultaneous optimization outperforms sequential approaches. The proposed WOA-SVR successfully eliminated 7 redundant features, reducing the dimensionality from 22 to 15, and achieved an MMRE of 21.32% ± 3.25% and a Pred (25) of 64.52% ± 3.90%. It significantly outperformed Standard SVR, PSO-SVR, GA-SVR, and GWO-SVR. Statistical validation via the Wilcoxon Signed-Rank Test and Cliff’s Delta effect size confirmed large and practically significant improvements. While the results demonstrate that simultaneous optimization provides a robust and simplified alternative for early-stage estimation, the effectiveness is bounded to the NASA93 dataset, necessitating future validation on modern agile repositories.
Construction of a Dialect-Sensitive Javanese Semantic Lexicon to Support Machine Translation Systems Musthofa Galih Pradana; Ridwan Raafi'udin; Nurul Afifah Arifuddin; Mohammad Asaduzzaman Rasel
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.v10i4.13379

Abstract

The development of linguistic resources for natural language processing (NLP) in Javanese remains limited, especially regarding the representation of semantic relationships between different speech levels. This study aims to construct a Javanese semantic lexicon that integrates Indonesian lexical equivalents with three Javanese speech levels: ngoko, krama alus, and krama inggil. A research design based on lexical resource construction was employed, using a Javanese digital dictionary as the primary data source. The methodology included data extraction, preprocessing, semantic lexicon construction, analysis of speech level variation, and a preliminary exploration of polysemous lexical entries using automatic identification, followed by validation by native speakers. The resulting semantic lexicon successfully represents lexical relationships between levels in a structured manner. Analysis of speech-level variation revealed that partially distinct lexical patterns were the most dominant, with 733 entries, followed by fully distinct patterns (193 entries) and identical patterns (21 entries). These findings indicate that speech-level differences in Javanese are selectively realized and should be explicitly considered in the development of linguistic resources. Furthermore, preliminary exploration of polysemous candidates demonstrated that dictionary-based automatic identification can overestimate polysemy without linguistic validation. Only a limited number of lexical entries exhibited features consistent with genuine polysemous relationships. This study provides an initial basis for the development of Javanese semantic resources that are sensitive to speech-level variation and semantic complexity. The constructed semantic lexicon has the potential to support future research in NLP applications in Javanese, including politeness identification, lexical normalization, word sense disambiguation, and machine translation.
Ensemble Tree-Based Machine Learning for Predicting Volumetric Change in Lithium-Based Battery Electrode Materials Arsenio Farrell Winoto; Gustina Alfa Trisnapradika
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.v10i4.13386

Abstract

Volumetric instability in lithium-based electrode materials remains a persistent challenge in electric vehicle battery development, as identifying stable material combinations through conventional laboratory methods is both time-consuming and resource-intensive. This study develops and compares three ensemble tree-based machine learning models, namely Random Forest, XGBoost, and CatBoost, to predict the maximum volume change percentage of lithium-based electrode materials. A dataset of 52,503 samples was constructed by integrating electrode pair data with structural and electronic features from the Materials Project API, enriched with compositional descriptors extracted using the Matminer Magpie preset. Each model underwent baseline evaluation followed by hyperparameter tuning using Optuna with Bayesian optimization over 100 trials, assessed using RMSE, MAE, and R². All three models achieved R² test above 0.98, with Random Forest yielding the best performance at RMSE of 17.1713, MAE of 4.0954, and R² test of 0.9900. SHAP analysis identified density discharge as the most determinant predictor across all models, reflecting its physicochemical role in representing the final crystal structure state following lithium intercalation. These findings confirm that ensemble tree-based models offer a reliable and efficient alternative to wet laboratory experimentation for lithium-based electrode material discovery.
TikTok Sentiment Analysis on Koperasi Merah Putih Using SVM and ANN Meliysa Pasa Bagna Aprilia Said; Kholifatus Sholihah; Ifnu Wisma Dwi Prastya; Afril Efan Pajri
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.v10i4.13394

Abstract

TikTok has become a relevant social media source for observing public responses to public issues, including the Koperasi Merah Putih program. This study compares Support Vector Machine (SVM) and Artificial Neural Network (ANN) for classifying sentiment in TikTok comments. The dataset was obtained through TikTok comment scraping and consisted of 25,669 raw comments. After removing empty comments and applying preprocessing stages consisting of cleaning, case folding, normalization, tokenization, stopword removal, and stemming, 21,026 comments were used for sentiment analysis. Sentiment labels were generated automatically using a lexicon-based sentiment labeling approach and grouped into three classes: positive, negative, and neutral. TF-IDF was used for feature extraction with a maximum of 5,000 features and unigram-bigram representation. The dataset was split into training and testing sets with an 80:20 ratio, while Stratified K-Fold Cross Validation and SMOTE were applied to strengthen evaluation and address class imbalance. The results show that SVM achieved the best overall performance before SMOTE with an accuracy of 86.66% and an F1-score of 86.76%. ANN achieved an accuracy of 85.31% before SMOTE and improved slightly after SMOTE to 85.47%. These findings indicate that SVM is more stable for TF-IDF-based TikTok comment classification, while SMOTE can improve ANN performance slightly but does not always increase all models equally.
Comparison of the Performance of K-Nearest Neighbor and Naive Bayes Algorithms for Sentiment Analysis of PinjamYuk Application User Reviews Using SMOTE and TF-IDF Lindya Rossita Handoko; Amelia Faza; Mula Agung Barata; Afril Efan Pajri
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.v10i4.13398

Abstract

The growth of online lending services has driven the increasing adoption of digital financial applications, providing users with convenient access to financial services. This growing number of users has generated a large volume of reviews on the Google Play Store, which can serve as a valuable source for understanding users’ perspectives on the benefits and performance of these applications. This study aims to compare the performance of the K-Nearest Neighbor (KNN) and Naive Bayes algorithms in classifying the sentiment of user reviews of the PinjamYuk application. The study uses a secondary dataset obtained from Kaggle, consisting of 500 user reviews of the PinjamYuk application on the Google Play Store during the 2023–2024 period. The reviews were categorized into three sentiment classes: positive, neutral, and negative, based on their rating scores. Because the class distribution in the dataset was imbalanced, the Synthetic Minority Oversampling Technique (SMOTE) was applied to balance the classes before the classification process. The research procedure included data preprocessing, feature extraction using Term Frequency–Inverse Document Frequency (TF-IDF), data balancing using SMOTE, and KNN parameter optimization using GridSearchCV. The models were evaluated using accuracy, precision, recall, F1-score, confusion matrix, stratified K-fold cross-validation, and the McNemar test. The results show that the Naive Bayes algorithm outperformed KNN. The stratified K-fold cross-validation results yielded an average accuracy of 81.0% for Naive Bayes and 80.8% for KNN. Furthermore, the McNemar test produced a p-value of 0.014 (p < 0.05), indicating that the performance difference between the two algorithms was statistically significant. These findings demonstrate that the Naive Bayes algorithm is more effective for analyzing user sentiment in reviews of the PinjamYuk application.
Comparison of LSTM, GRU, Bi-LSTM, and XGBoost for BRIS Stock Price Prediction with Lookback Period Variations Subekti Wahyu Aji; Wise Herowati
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.v10i4.13412

Abstract

Stock price forecasting remains a challenging task due to the nonlinear and non-stationary characteristics of financial time series, particularly for Islamic banking stocks such as Bank Syariah Indonesia (BRIS), which exhibit highly dynamic price movements. This study compares the performance of four prediction models Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Bidirectional Long Short-Term Memory (Bi-LSTM), and XGBoost in forecasting BRIS closing stock prices using OHLCV data collected from August 2020 to December 2025. To investigate the influence of historical observation windows, three lookback periods (10, 20, and 30 trading days) were evaluated. Model performance was assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R²). The experimental results demonstrate that the GRU model with a 10-day lookback period achieved the best overall performance, yielding an RMSE of 78.0643 IDR, a MAPE of 2.13%, and an R² value of 0.8438. It outperformed Bi-LSTM, XGBoost, and LSTM, all of which also achieved their highest accuracy using the 10-day lookback period. Furthermore, the analysis reveals that the shorter lookback window consistently provides superior predictive performance across all evaluated models, whereas extending the lookback period to 20 or 30 days does not lead to consistent performance improvements. Residual analysis further indicates a relatively unbiased distribution of prediction errors. These findings suggest that the simplified gating mechanism of GRU enables better stability and generalization than more complex architectures for forecasting Islamic bank stock prices. This study contributes to the understanding of the relationship between model complexity and historical time window selection in deep learning-based financial forecasting and provides practical guidance for selecting appropriate prediction models for Indonesian Islamic banking stocks.
Work Readiness Prediction of Vocational High School Students Using a Genetic Algorithm-Optimized Random Forest and XGBoost as a Benchmark Model with Explainable AI (SHAP) Syifa Amalia; Noor Latifah; Andy Prasetyo Utomo
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.v10i4.13435

Abstract

Work readiness among vocational high school (SMK) students remains a critical challenge in aligning educational outcomes with industry expectations. This study proposes a binary classification system using Genetic Algorithm (GA)-optimized Random Forest as the primary model and XGBoost as a benchmark. Two predictor variables are used: average practical score (X1) and attitude score (X2). The classification target is based on the minimum competency threshold (KKM = 80) regulated by the Indonesian Ministry of Education: a student is labeled Ready to Work when X1 >= 80 AND X2 >= 80. A boundary noise injection mechanism is applied to mitigate label determinism at the borderline zone. The dataset comprises 1,211 student records split into 968 training and 243 testing records. Evaluation employs hold-out testing, Stratified 5-Fold Cross Validation, overfitting detection, and McNemar's statistical significance test. SHAP interpretability is performed at three levels: global feature importance, beeswarm summary plot, and dependence plots. RF+GA achieved 90.12% accuracy, 0.9391 precision, 0.8640 recall, 0.9000 F1-score, and 0.9582 AUC-ROC. Cross-validation confirmed stable performance at 91.53% +/- 2.51%. Overfitting analysis showed a training-testing gap of only 1.72% (Not Detected). McNemar's test (p = 1.000) indicated statistically equivalent performance between both models. SHAP analysis revealed that attitude score (X2) is the dominant predictor (mean SHAP = 0.2611), followed by average practical score (X1, mean SHAP = 0.1628). Practical implications include an early warning system, targeted remedial programs, and individual guidance recommendations generated by the deployed Streamlit application.
OCR Engines for License Plate Recognition: A Comparative Study of Tesseract, EasyOCR, PaddleOCR, and TrOCR Afifah Khaerani Aziz; Marzuki Pilliang; Diana Kuniawati
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.v10i4.13437

Abstract

Automatic License Plate Recognition (ALPR) is a critical component of Intelligent Transportation Systems (ITS); however, the Optical Character Recognition (OCR) stage remains a significant bottleneck when confronting diverse linguistic scripts, complex plate formats, and environmental degradations. This systematic literature review comparatively evaluates four primary OCR engines—Tesseract, EasyOCR, PaddleOCR, and TrOCR—to bridge the gap between constrained local optimizations and globally resilient applications. Adhering to the PRISMA 2020 guidelines, a comprehensive search across six major academic databases spanning January 2020 to May 2026 initially identified 647 records. Following rigorous screening processes, a final core corpus of 21 empirical studies was qualitatively synthesized to account for extreme cross-study hardware and dataset heterogeneity. The analysis reveals that no single engine is universally superior; efficacy is fundamentally dictated by their underlying neural architectures and contextual deployment parameters. Tesseract offers maximum computational efficiency but fails significantly on non-Latin and complex scripts due to legacy segmentation limits. EasyOCR provides an optimal accuracy-to-speed ratio, making it highly suitable for real-time edge device deployments. PaddleOCR excels in robust sequential decoding, delivering the highest exact-match rates required for high-stakes applications like automated tolling. Conversely, the Transformer-based TrOCR emerges as the definitive frontier for highly complex, irregularly spaced, and multilingual plates (e.g., Han, CIS region scripts), though its severe computational latency currently restricts it to cloud-based infrastructures. Furthermore, the synthesis establishes that dynamic, environment-aware preprocessing is a mandatory prerequisite to mitigate visual stressors such as motion blur and low illumination. This review provides a novel, context-driven deployment taxonomy, equipping researchers and developers with actionable guidelines to navigate the trade-offs between architectural accuracy, script diversity, and computational resource constraints. Finally, the review identifies unified Large Vision-Language Models (LVLMs) as the next-generation trajectory to resolve current multi-stage pipeline dependencies.

Page 96 of 101 | Total Record : 1006

 94   95   96   97   98 

Filter by Year

2017 2026


Reset
Filter By Issues
All Issue Vol. 10 No. 4 (2026): August 2026 Vol. 10 No. 3 (2026): June 2026 Vol. 10 No. 2 (2026): April 2026 Vol. 10 No. 1 (2026): February 2026 Vol. 9 No. 6 (2025): December 2025 Vol. 9 No. 5 (2025): October 2025 Vol. 9 No. 4 (2025): August 2025 Vol. 9 No. 3 (2025): June 2025 Vol. 9 No. 2 (2025): April 2025 Vol. 9 No. 1 (2025): February 2025 Vol. 8 No. 2 (2024): December 2024 Vol. 8 No. 1 (2024): July 2024 Vol. 7 No. 2 (2023): December 2023 Vol. 7 No. 1 (2023): July 2023 Vol. 6 No. 2 (2022): December 2022 Vol 6 No 1 (2022): July 2022 Vol. 5 No. 2 (2021): December 2021 Vol 5 No 2 (2021): December 2021 Vol 5 No 1 (2021): July 2021 Vol. 5 No. 1 (2021): July 2021 Vol 4 No 2 (2020): Desember 2020 Vol. 4 No. 2 (2020): Desember 2020 Vol 4 No 1 (2020): Juli 2020 Vol 3 No 2 (2019): Desember 2019 Vol 3 No 1 (2019): Juli 2019 Vol 2 No 2 (2018): Desember 2018 Vol. 2 No. 1 (2018): Juli 2018 Vol 2 No 1 (2018): Juli 2018 Vol 1 No 2 (2017): Desember 2017 Vol 1 No 1 (2017): Juli 2017 More Issue