cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
Suyahman
Contact Email
suyahman.com@gmail.com
Phone
+6285155377406
Journal Mail Official
sahpublisher@gmail.com
Editorial Address
Jl. Pedan Karangdowo, Munggung, Karangdowo, Klaten, Central Java, Indonesia
Location
Kab. klaten,
Jawa tengah
INDONESIA
Journal of Artificial Intelligence and Legal Technology
Published by PT Solusi Administrasi Hukum
ISSN : -     EISSN : 3123786X     DOI : -
Core Subject :
The Journal of Artificial Intelligence and Legal Technology (JAILT) is an international, peer-reviewed journal dedicated to advancing interdisciplinary research in artificial intelligence (AI) and its applications in the legal domain. JAILT serves as a platform for academics, practitioners, and policymakers to discuss cutting-edge developments in AI-driven legal reasoning, regulatory compliance, and intelligent legal systems. The journal publishes high-quality research articles, reviews, and research notes that address both theoretical and applied aspects of AI in law, fostering dialogue on legal, ethical, and societal challenges.
Arjuna Subject : -
Articles 10 Documents
 1 
Comparative Analysis of Trademark Class Identification Using IndoBERT and Multilingual BERT Ayun Hapsari; Suyahman Suyahman
Journal of Artificial Intelligence and Legal Technology Vol. 1 No. 1 (2025): August 2025
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

The rapid growth of trademark registrations in Indonesia has increased the demand for efficient and accurate classification into the internationally recognized NICE system. Manual assignment of classes remains time-consuming and prone to human error, motivating the need for an automated approach. This study investigates the application of Transformer-based language models for trademark class identification based solely on product and service descriptions. Two models were evaluated: the Multilingual BERT (mBERT) and the monolingual IndoBERT, both fine-tuned for sequence classification across 45 NICE classes using 59,948 trademark entries collected from the Directorate General of Intellectual Property (DGIP) database. The research methodology encompassed data preprocessing, stratified train-test splitting (80:20), and tokenization with a maximum sequence length of 64 tokens. Both models were trained for two epochs using the AdamW optimizer, and evaluated with accuracy, precision, recall, F1-score, and per-class accuracy (one-vs-all). Experimental results reveal that IndoBERT significantly outperforms mBERT, achieving an overall accuracy, precision, recall, and F1-score of 0.90, compared to 0.85 for mBERT. IndoBERT demonstrated particularly robust performance in low-support classes, indicating its superior ability to capture domain-specific linguistic nuances in Indonesian trademark descriptions. The findings underscore the potential of monolingual Transformer models for automating trademark classification in national intellectual property systems. The integration of such models can accelerate trademark registration, reduce examiner workload, and enhance consistency in class assignment. These results contribute to advancing the deployment of AI in legal and administrative contexts, while providing a foundation for future work involving multimodal features and explainable AI for comprehensive trademark management solutions.
Semantic Analysis of Trademark Names Using Large Language Models Muhammad Adi Pratama; Suyahman Suyahman
Journal of Artificial Intelligence and Legal Technology Vol. 2 No. 1 (2026): February 2026
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

This study introduces a novel framework for trademark similarity analysis that integrates large language models (LLMs) to assess lexical, phonetic, and semantic relationships between trademark names without reliance on large precompiled databases or retraining. The primary motivation is to address the growing need for efficient and transparent preliminary trademark screening, which is often constrained by the limitations of traditional rule-based or string-matching approaches. To achieve this, a web-based system was developed using the Gemini API, allowing users to input trademark pairs for automated analysis. The workflow includes text normalization, phonetic conversion, multi-dimensional similarity computation, and the generation of interpretative explanations for each pair. A test dataset of ten diverse trademark name pairs was designed to capture variations in lexical overlap, phonetic similarity, and semantic association. The system’s outputs were evaluated both in terms of processing efficiency and expert assessment. Quantitative results show that the system can process a pair in under a second on average, handling 3,229 tokens across ten pairs with minimal computational overhead. Qualitative evaluation by five trademark and intellectual property experts using a 5-point Likert scale yielded mean scores of 4.4 for relevance, 3.8 for explanation quality, and 4.2 for practical usefulness, confirming strong alignment between the LLM outputs and expert intuition. The novelty of this research lies in demonstrating that LLMs can provide not only accurate similarity assessments but also human-readable interpretative reasoning, bridging the gap between automation and expert judgment in trademark evaluation. This approach offers a transparent and scalable solution for early-stage brand screening, significantly reducing the reliance on extensive databases and manual effort. The findings indicate a clear potential for integration into industrial-scale trademark examination workflows, paving the way for future developments in batch processing, recommendation systems, and enhanced interpretability in AI-assisted intellectual property management.
Optimized Machine Learning with TPE for Air Quality Classification and Public Health Risk Estimation Ayun Hapsari
Journal of Artificial Intelligence and Legal Technology Vol. 1 No. 1 (2025): August 2025
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Air pollution in rapidly urbanizing cities such as Delhi poses a critical threat to public health due to the persistent exceedance of safe thresholds for particulate matter and gaseous pollutants. Accurate air quality classification and timely health risk estimation are essential to support early warning systems and guide urban policy interventions. This study develops a multi-class Air Quality Index (AQI) classification framework using Logistic Regression, Random Forest, Decision Tree, Support Vector Classifier (SVC), K-Nearest Neighbors (KNN), and Gradient Boosting, applied to a comprehensive dataset of daily pollutant concentrations (PM2.5, PM10, NO₂, SO₂, CO, and O₃) and meteorological parameters from Delhi. Data preprocessing included outlier removal, feature scaling, and label encoding of AQI categories, followed by an 80:20 train-test split to ensure robust model evaluation. Model performance was assessed using Accuracy, Precision, Recall, and F1-score. The experimental results show that ensemble and kernel-based models achieved the highest predictive accuracy, with Random Forest reaching an accuracy of 0.7611 and an F1-score of 0.7522, followed closely by Decision Tree and Gradient Boosting with F1-scores above 0.74. Logistic Regression and SVC maintained moderate yet consistent performance, while KNN was more sensitive to data distribution, achieving an F1-score of 0.605. Confusion matrix analysis revealed that misclassifications were mostly confined to adjacent AQI categories, reflecting the natural difficulty of distinguishing borderline pollution levels. The novelty of this study lies in integrating multi-class AQI classification with a structured machine learning framework capable of mapping environmental conditions directly to health risk levels. By aligning predictions with WHO and US-EPA thresholds, the framework facilitates actionable insights for public health authorities, enabling the design of early warning systems and targeted interventions for vulnerable populations. These findings advance the technical landscape of urban air quality management and provide a scalable foundation for health-oriented environmental decision-making in highly polluted megacities.
Data Analysis and Explainable Machine Learning for Stunting Prediction Ardy Wicaksono; Deny Prasetyo; Yulaikha Mar'atullatifah; Dwi Utari Iswavigra; Himmatunnisak Mahmudah; Ayun Hapsari
Journal of Artificial Intelligence and Legal Technology Vol. 1 No. 1 (2025): August 2025
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Childhood stunting remains a critical global health concern, reflecting chronic malnutrition that affects both physical growth and long-term cognitive development. Despite ongoing interventions, early detection in many low- and middle-income countries is still hindered by limited resources and the absence of interpretable decision-support tools. This study aims to develop and evaluate an explainable machine learning framework to predict stunting among toddlers using simple anthropometric and demographic data, thereby supporting evidence-based public health interventions. Data were collected from 40,071 children aged 0–59 months in Jeneponto Regency, South Sulawesi, Indonesia, covering the period 2021–2024. Key features included age in months, gender, weight, and height, while stunting status served as the target variable. Several machine learning algorithms were implemented, including Logistic Regression, Support Vector Machine, Multilayer Perceptron, K-Nearest Neighbors, Decision Tree, Random Forest, XGBoost, and Convolutional Neural Network. Data preprocessing involved imputation of missing values, feature encoding, and an 80/20 train-test split, while model interpretability was achieved using SHAP (SHapley Additive exPlanations) to provide both global and local feature attributions. The experimental results show that XGBoost achieved the highest accuracy of 97.57%, followed closely by Random Forest (97.28%) and Decision Tree (96.62%). SHAP analysis revealed that height was the most influential predictor, followed by age, gender, and weight, providing actionable insights for early identification of at-risk children. Local SHAP force plots further enabled case-level interpretation, enhancing the trustworthiness of the model in clinical or community health applications. The novelty of this research lies in integrating high-performing machine learning models with explainable AI for stunting prediction using minimal, easily collected health features in a resource-limited context. This framework not only improves the accuracy and transparency of early stunting detection but also provides a scalable approach to strengthen nutrition surveillance systems, with potential to inform targeted interventions and reduce the long-term impacts of childhood malnutrition.
Classification of Stock Listing Boards for Warrants using Machine Learning and Bayesian Optimization Deny Prasetyo; Muhammad Anwar Fauzi
Journal of Artificial Intelligence and Legal Technology Vol. 1 No. 1 (2025): August 2025
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Automatic classification of warrant stock listing boards is an important challenge in managing capital market information, especially on the Electronic Indonesia Public Offering (E-IPO) platform. This research implements various machine learning algorithms optimized using Bayesian Optimization to improve the classification accuracy of six listing board categories. Ensemble models such as Random Forest, CatBoost, and XGBoost showed superior performance with the highest accuracy reaching 74.68%. The use of Bayesian Optimization effectively finds the optimal hyperparameters, strengthening the overall performance of the model. Evaluation was conducted through stratified cross-validation and confusion matrix analysis, providing in-depth insight into prediction accuracy. The results of this research contribute to the automation of listing board clustering that supports the strategic decisions of investors, issuers, and regulators in the Indonesian capital market.
An Analysis of the Impact of Social Media Addiction on Students’ Academic Performance Using K-Means and Decision Tree Dewi Sayekti Sutrisni; Maulana Ilham Alisyahbana; Muhammad Luqman Al-hakim; Deny Prasetyo
Journal of Artificial Intelligence and Legal Technology Vol. 1 No. 1 (2025): August 2025
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

This study aims to analyze the relationship between social media addiction levels and students' academic performance. With the growing use of social media among university students, concerns have emerged regarding its potential negative impact on academic achievement. The data were obtained from the "Social Media Addiction vs Relationships" dataset and analyzed using two machine learning approaches: K-Means to classify groups based on usage hours and academic impact, and Decision Tree to predict academic satisfaction levels based on digital behavior. The findings reveal distinct clustering patterns that differentiate students based on their addiction levels and academic performance. The Decision Tree model achieved 100% accuracy on the test data in classifying the impact of social media use. These results highlight that daily usage hours and addiction scores are significant contributing factors. Based on these insights, the study recommends implementing digital intervention programs on campus to help mitigate the negative effects of social media addiction.
Performance Comparison of Supervised Classifiers in Intrusion Detection Systems Onesinus Tamba
Journal of Artificial Intelligence and Legal Technology Vol. 2 No. 1 (2026): February 2026
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Intrusion Detection System (IDS) is an essential component in network security to detect and respond to cyber attacks. This study explores the use of several supervised classifier algorithms to classify attacks using the KDDTest-21 dataset from NSL-KDD. This dataset was chosen because of its improvement over the KDD'99 dataset that reduces bias due to duplication and uneven data distribution. The algorithms used include Logistic Regression, KNearest Neighbors (KNN), Gaussian Naive Bayes, Support Vector Machine (SVM), Decision Treem, Random Fores  and Deep Feedforward Neural Network. Each algorithm is applied to a normalized dataset using RobustScaler. Performance assessment is carried out based on metrics such as accuracy, precision, recall, and F1-score to determinethe best algorithm in detecting various types of attacks. The results show that several algorithms, especially Random Forest, have better performance in detecting attacks with high accuracy with 97.68% and other merics score are more than 98%. This finding is important for optimizing a more effective IDS in protecting network infrastructure from increasingly complex cyber attacks.
Trademark Logo Infringement Detection: A Threshold Determination Approach Using Bayesian Optimization and Siamese Neural Networks Muhammad Rifzal Alief Ramadhan; Stevano Wahyu Al'fandi; Egi Dio Bagus Sudewo; Suyahman Suyahman
Journal of Artificial Intelligence and Legal Technology Vol. 2 No. 1 (2026): February 2026
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Trademark infringement detection is essential to prevent consumer confusion, particularly in the digital era where visually similar logos are widespread. This study proposes a similarity threshold determination framework using Siamese Neural Networks (SNN) combined with Bayesian Optimization to improve the accuracy of trademark similarity assessment. Logo images were collected from the Indonesian intellectual property database, preprocessed into a uniform format, and trained using a triplet loss approach. Bayesian Optimization was applied to determine the optimal similarity threshold, minimizing false positive and false negative classifications. The proposed model achieved an accuracy of 92.23%, with precision of 93.34%, recall of 91.44%, and an F1-score of 92.39%. The optimal threshold (0.200149) effectively balanced sensitivity and specificity, resulting in low misclassification rates. These findings demonstrate that integrating SNN with Bayesian Optimization provides a robust and legally relevant framework for trademark infringement detection, offering practical implications for strengthening intellectual property protection.
An Experimental Evaluation of Machine Learning Models for Judicial Decision Prediction Using Indonesian Court Decisions Dybio Asih
Journal of Artificial Intelligence and Legal Technology Vol. 2 No. 1 (2026): February 2026
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Judicial outcome analysis has attracted growing attention within legal artificial intelligence research; however, empirical studies focusing on Indonesian court decisions remain limited. This study presents an experimental evaluation of traditional machine learning and deep learning models for judicial outcome classification using Indonesian legal texts.The experiments were conducted on a curated dataset of 4,872 court decisions obtained from the official Direktori Putusan Mahkamah Agung Republik Indonesia (2018–2023). To prevent outcome leakage, all explicit ruling sections were removed prior to model training, and only the legal reasoning segments were used as input. Several models, including Logistic Regression, Support Vector Machine, Gradient Boosting, BiLSTM, and IndoBERT, were evaluated under identical experimental settings. The results show that ensemble-based methods, particularly Gradient Boosting, achieve strong and stable performance, while deep learning models demonstrate competitive but not consistently superior results under document length constraints. Error analysis indicates that misclassifications frequently arise from implicit judicial reasoning and outcome ambiguity. This study provides an empirical benchmark for judicial outcome classification in Indonesian courts and highlights methodological limitations related to document length, labeling granularity, and reproducibility in legal NLP research.
Decision Support System for Selecting Social Media Platforms for Gen Z Personal Branding Using the TOPSIS Method Dyah Irmawati; Yulaikha Mar’atullatifah
Journal of Artificial Intelligence and Legal Technology Vol. 2 No. 1 (2026): February 2026
Publisher : Sah Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

The development of social media has encouraged Generation Z to utilize it as a primary means of building personal branding. However, the large number of social media platforms creates difficulties in determining the most appropriate and effective platform. This study aims to determine the best social media platform to support Generation Z's personal branding using a Decision Support System (DSS) approach with the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method. This study used a quantitative design with primary data obtained through questionnaires. The research respondents were Generation Z who actively use social media. The alternatives analyzed included five social media platforms evaluated based on seven assessment criteria relevant to personal branding needs. Data were analyzed using the TOPSIS method through the stages of compiling a decision matrix, normalization, weighting, determining positive and negative ideal solutions, calculating Euclidean distance, and determining preference values. The results showed that Instagram obtained the highest preference value of 0.5726 and was ranked first, followed by LinkedIn, TikTok, and YouTube. Meanwhile, alternative X (Twitter) had the lowest preference value, namely 0.2046. These findings demonstrate that the TOPSIS method is capable of providing an objective and systematic ranking of alternatives. This research contributes to providing a decision-making model that can be used as a practical reference for Generation Z in selecting optimal social media platforms for personal branding.

Page 1 of 1 | Total Record : 10

 1 

Filter by Year

2025 2026


Reset
Filter By Issues
All Issue Vol. 2 No. 1 (2026): February 2026 Vol. 1 No. 1 (2025): August 2025