Bulletin of Intelligent Machines and Algorithms
BIMA (Bulletin of Intelligent Machines and Algorithms) is an international peer-reviewed journal dedicated to promoting research in the fields of artificial intelligence, machine learning, and algorithms. BIMA serves as a platform for publishing the latest research findings and innovative applications in these rapidly evolving fields. The journal aims to contribute to the academic and professional development of researchers, practitioners, and educators by publishing high-quality articles that provide in-depth insights into the theoretical, practical, and computational aspects of intelligent systems and algorithms. Focus and Scope BIMA publishes original research articles, reviews, and technical reviews on various topics related to intelligent machines and algorithms. The scope of this journal includes, but is not limited to: Artificial Intelligence: Methodologies, algorithms, and architectures for building intelligent systems, including knowledge representation, reasoning, learning, and perception. Machine Learning: Supervised, unsupervised, semi-supervised, and reinforcement learning algorithms; applications in real-world problems. Deep Learning: Advanced neural network architectures such as CNNs, RNNs, Transformers, and their applications in various domains including image, video, text, and signal processing. Computer Vision: Image processing, object detection and recognition, image segmentation, motion analysis, and visual scene understanding in intelligent systems. Data Mining: Techniques for extracting patterns and knowledge from large datasets. Optimisation Algorithms: Theory and applications of optimisation techniques in continuous and discrete domains. Computational Intelligence: Evolutionary algorithms, fuzzy logic, and swarm intelligence systems. Natural Language Processing (NLP): Advances in language understanding, translation, and text analysis. Applications: Applications of artificial intelligence and algorithms in healthcare, finance, industry, education, and other fields. Robotics and Autonomous Systems: Intelligent robots, human-robot interaction, and autonomous vehicles.
Articles
25 Documents
<![CDATA[A Progressive Training Framework for Robust YOLOv11-Based Vehicle Detection Across Domain Shifts in Real-World CCTV Environments ]]>
<![CDATA[Aditya Gunaldhi]]>;
<![CDATA[Budiman]]>;
<![CDATA[Chairul Habibi]]>;
<![CDATA[Nur Alamsyah]]>
<![CDATA[ Bulletin of Intelligent Machines and Algorithms Vol. 1 No. 5 (2026): BIMA July 2026 Issue ]]>
Publisher : <![CDATA[Maheswari Publisher ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.65780/bima.v1i5.27
<![CDATA[The performance of deep learning-based vehicle detection models often deteriorates when applied to real-world CCTV environments due to domain shift caused by variations in lighting, occlusion, glare, and changes in camera viewpoint. This study aims to develop a YOLOv11-based progressive training framework to improve the model’s generalization ability under heterogeneous operational conditions. The proposed method consists of four stages: base training, fine-tuning, and a supervised progressive domain adaptation strategy implemented through sequential fine-tuning on labeled target-domain CCTV images before OpenVINO-based inference optimization. The model is trained using a source dataset and adapted to a target dataset representing real-world CCTV conditions. Evaluation was conducted using Precision, Recall, mAP@50, mAP@50–95, loss curve analysis, per-vehicle-class evaluation, visual testing on CCTV video, as well as latency and throughput measurements. The results show that base training achieved an mAP@50 of 0.919 and built a robust feature representation, while fine-tuning maintained performance stability with an mAP@50 of 0.915. Although domain adaptation reduced mAP@50 to 0.811 due to domain shift, the model demonstrated improved generalization capabilities and maintained more consistent detection under low-light conditions, glare, occlusion, and heavy traffic. OpenVINO INT8 optimization increased inference speed from 4.86 FPS to 9.08 FPS with minimal accuracy loss. These findings demonstrate that the progressive training framework effectively bridges differences in data distribution while producing a vehicle detection model that is more robust, efficient, and suitable for deployment in real-time CCTV-based traffic monitoring systems.]]>
<![CDATA[Content-Based Filtering with TF-IDF and Cosine Similarity for Bandung Raya Tourism Recommendation ]]>
<![CDATA[Mira Aldina]]>;
<![CDATA[Galih]]>;
<![CDATA[Siti Nur]]>
<![CDATA[ Bulletin of Intelligent Machines and Algorithms Vol. 1 No. 5 (2026): BIMA July 2026 Issue ]]>
Publisher : <![CDATA[Maheswari Publisher ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.65780/bima.v1i5.28
<![CDATA[Bandung Raya offers hundreds of tourist destinations spread across Bandung City, Bandung Regency, and West Bandung Regency, which causes information overload and makes it difficult for tourists to select destinations that match their preferences. Recommender systems, as one of the most widely applied branches of machine learning, provide a way to filter such information automatically. This study develops a machine learning-based recommender system for tourism destinations in Bandung Raya using Content-Based Filtering, in which destination descriptions are represented as numerical vectors through TF-IDF weighting and compared using Cosine Similarity. The dataset consists of 331 destinations with name, description, category, and region attributes. Text preprocessing is performed in five stages using the Sastrawi library for the Indonesian language, producing a TF-IDF matrix of 331 by 583 and a similarity matrix of 331 by 331. The model is deployed as a website using Flask as the backend, React as the frontend, and a REST API as the interface, supporting both name-based search and free-text query search. Functional validation uses Black-Box Testing, while recommendation quality is measured using Precision at K and Mean Average Precision. All eight functional scenarios passed, with Precision at 5 of 96.00 percent, Precision at 10 of 90.00 percent, and Mean Average Precision of 98.86 percent, indicating that, for the five evaluated queries, relevant destinations are ranked highly.]]>
<![CDATA[A Trigger Aware, Event Centric, and Uncertainty Calibrated Neuro-Symbolic Framework for Actionable Cyber Threat Intelligence from Indonesian Online News ]]>
<![CDATA[Elia Setiana]]>;
<![CDATA[Muhamad Achya Arifudin]]>;
<![CDATA[Nur Alamsyah]]>
<![CDATA[ Bulletin of Intelligent Machines and Algorithms Vol. 1 No. 5 (2026): BIMA July 2026 Issue ]]>
Publisher : <![CDATA[Maheswari Publisher ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.65780/bima.v1i5.31
<![CDATA[Online news can provide timely cyberthreat signals, but duplicative reporting, fragmented event descriptions, resource-constrained Indonesian language text, and uncalibrated model confidence limit its operational use. This study presents A Trigger-Aware, Event-Centric, and Uncertainty-Calibrated Neuro-Symbolic Framework for Actionable Cyber Threat Intelligence from Indonesian Online News (TRACE-CTI-ID), a proof-of-concept framework that integrates exact deduplication, event-centric clustering, trigger-aware semantic representation, neuro-symbolic fusion, ordinal risk estimation, conformal uncertainty, mitigation mapping, and an event-centric knowledge graph. The experiment used 711 Liputan6 records collected on March 14, 2025. Exact deduplication reduced the corpus to 79 unique headlines, which were automatically consolidated into 26 events. Splitting the separate events resulted in 30 training articles, 5 calibration articles, and 44 test articles with zero event leakage. The calibrated neuro-symbolic model achieved a micro-F1 of 0.283 and a macro-F1 of 0.441, outperforming the baseline TF-IDF of 0.074 and 0.013, respectively. However, ordinal severity prediction remained weak with an accuracy of 0.136, a macro-F1 of 0.138, and a mean absolute error of 2.023. Conformal coverage was also unstable, and the abstention mechanism did not direct uncertain articles to human review. These findings demonstrate the technical feasibility of the integrated pipeline while also demonstrating that silver labeled, title only, and single source data are insufficient for final operational validation. Therefore, the key contribution is a transparent, leak aware evaluation architecture and protocol that can be strengthened through full text collection from multiple sources and independent expert annotation.]]>
<![CDATA[Intelligent Cold Chain Control System Based on Data Fusion for Refrigerated Truck Fleets ]]>
<![CDATA[Luckman Hasannudin]]>;
<![CDATA[Very Kurnia Bakti]]>;
<![CDATA[Abdul Basit]]>
<![CDATA[ Bulletin of Intelligent Machines and Algorithms Vol. 1 No. 5 (2026): BIMA July 2026 Issue ]]>
Publisher : <![CDATA[Maheswari Publisher ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.65780/bima.v1i5.32
<![CDATA[As a maritime nation, Indonesia holds substantial marine-resource potential, yet its distribution faces serious challenges because fishery commodities are highly perishable and prone to rapid quality decline. Failure to maintain consistent temperatures throughout the cold chain frequently causes significant economic losses for business operators. This study aims to design and build an Intelligent Cold Chain Control System Based on Data Fusion for Refrigerated Truck Fleets using an ESP32-S3 microcontroller. The system integrates Internet of Things (IoT) technology with a data fusion technique that combines readings from an RTD PT100 temperature sensor paired with a MAX31865 module, an MPU6050 vibration sensor, and a NEO-8M GPS module for real-time location tracking, using an SD card as a buffer medium and the MQTT protocol for transmitting data to a server. The research method applied is the Prototype method, comprising requirements analysis, initial system design, prototype development, user evaluation, prototype improvement, prototype testing, and implementation. Data were collected through field observation, interviews, and a literature study at CV. Mutiara Samudera Indonesia, Tegal. Test results show that the PT100 sensor was able to read temperatures down to −15.5°C with a deviation of only 0.2°C against a reference instrument, the NEO-8M GPS module successfully displayed location coordinates within a tolerance of a few meters compared with Google Maps, and the MPU6050 sensor was able to detect vibration changes on three axes in a stable manner. This prototype is expected to support quality monitoring and distribution efficiency for marine commodities carried by refrigerated truck fleets.]]>
<![CDATA[Topic Modelling of TikTok Application User Complaints in Google Play Store Reviews Using Latent Dirichlet Allocation ]]>
<![CDATA[Ghevira Devanda]]>;
<![CDATA[Inaya Zehan Kalyzta]]>;
<![CDATA[Zaqi Kurniawan]]>
<![CDATA[ Bulletin of Intelligent Machines and Algorithms Vol. 1 No. 5 (2026): BIMA July 2026 Issue ]]>
Publisher : <![CDATA[Maheswari Publisher ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.65780/bima.v1i5.33
<![CDATA[TikTok is a short-video-sharing platform with an extremely large user base in Indonesia, and as a result it also receives numerous negative reviews concerning both technical performance and platform policies on the Google Play Store. This study aims to identify and cluster the main topics of TikTok user complaints using a Machine Learning-based approach built on the Latent Dirichlet Allocation (LDA) algorithm. Data were collected through a quota-based web scraping procedure using the google-play-scraper library, targeting Indonesian-language reviews with a rating below three (rating 1 and 2) up to a target of 1,000 reviews, yielding a raw sample of 1,000 reviews (814 one-star and 186 two-star). The preprocessing stage included cleaning, case folding, normalization using a colloquial lexicon enriched with a TikTok domain-specific correction dictionary, stopword removal, and stemming using Sastrawi, which reduced the data to 814 clean documents. Document representation was constructed using a Bag of Words (BoW) approach through the Gensim library, with vocabulary filtering (no_below = 3, no_above = 0.6) that produced a dictionary consisting of 397 unique tokens. The optimal number of topics was determined through a multi-metric evaluation combining Coherence Score (u_mass), Coherence Score (c_v), and Perplexity for candidate models ranging from two to five topics, complemented by a manual check of topic interpretability. Despite mixed signals from the c_v and Perplexity metrics, the two-topic model was selected based on its u_mass score (-8.7713) and its more interpretable, non-fragmented thematic structure. The document distribution results show that Topic 1 is more dominant, comprising 445 documents (54.67%), representing technical application complaints such as bugs, slow performance, failure to open the application, and excessive advertisements. Meanwhile, Topic 2, with 369 documents (45.33%), relates to complaints about content moderation and restrictions on the live streaming feature. Visualization using pyLDAvis confirmed that the two topics are clearly separated without overlap, reinforcing the validity of the modelling results. As the sample is restricted to Indonesian-language, one- and two-star reviews collected at a single point in time, the findings are intended to reflect this specific segment of TikTok users rather than the full population of TikTok reviews. This study provides objective insights for application developers in prioritizing system improvements and evaluating content moderation policies more effectively.]]>
