cover
Contact Name
Mustakim
Contact Email
Mustakim
Phone
-
Journal Mail Official
ijaidm@uin-suska.ac.id
Editorial Address
-
Location
Kab. kampar,
Riau
INDONESIA
Indonesian Journal of Artificial Intelligence and Data Mining
ISSN : 26143372     EISSN : 26146150     DOI : -
Core Subject : Science,
Indonesian Journal of Artificial Intelligence and Data Mining (IJAIDM) is an electronic periodical publication published by Puzzle Research Data Technology (Predatech) Faculty of Science and Technology UIN Sultan Syarif Kasim Riau, Indonesia. IJAIDM provides online media to publish scientific articles from research in the field of Artificial Intelligence and Data Mining. IJAIDM will be published 2 (two) times a year, in March and September, each edition contains 7 (seven) articles. Articles may be written in English or Indonesia.
Arjuna Subject : -
Articles 250 Documents
Comparative Study: Performance Comparison of You Only Look Once and Convolutional Neural Networks Algorithms in Human Object Detection Sari, Dewi Permata; Ramadhani, M. Akbar Tri; Abdurrahman, Abdurrahman
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i3.37676

Abstract

The evolution of object identification technologies, particularly for person detection applications, has increasingly accelerated due to the merger of deep learning and artificial intelligence with computer vision. This study intends to test the efficacy of two object detection algorithms, YOLOv8n and CNN MobileNetSSD, in identifying human objects in digital photos. A dataset of 12,334 human-labeled photos from the Roboflow platform was utilized to train the YOLOv8n model, while performance results for the CNN MobileNetSSD model were acquired from a prior article. The precision, recall, and F1-score of each model were examined. Experimental results reveal that YOLOv8n attains 94% precision, 92% recall, and a 92.9% F1-score, representing a considerable enhancement over MobileNetSSD. Conversely, MobileNetSSD got an F1-score of 85.2%, with a precision of 86.5% and a recall of 84.1%. The findings show that CNN MobileNetSSD is more ideal for non-time-sensitive or resource-limited scenarios; however, YOLOv8n is preferable for real-time human identification tasks due to its greater accuracy and faster inference. This comparative analysis is important for differentiating object detection models matched to certain application needs.
Acne Skin Detection System Using You Only Look Once (YOLOV8) Based on Artificial Intelligence Sabara, Gally; Abdurrahman, Abdurrahman; Sari, Dewi Permata; Kurniawan, Aprila
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 2 (2025): July 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i2.37217

Abstract

Acne is one of the most common skin problems among teenagers and young adults, and early detection is essential to prevent progression and long-term skin damage. This study aims to develop a real-time acne detection system utilizing the YOLOv8 deep learning algorithm, integrated with a Raspberry Pi and webcam, and supported by Telegram-based notifications for user monitoring. The dataset comprises 4,092 annotated facial images representing three types of acne: papule, pustule, and nodule. Model training was conducted in Google Colab with appropriate hyperparameter adjustments. The evaluation results show that the model performs well in detecting papule and pustule acne types, with correct predictions of 258 and 222 samples, respectively, in the confusion matrix, although misclassification remains high for comedones and background classes. The Precision–Confidence Curve indicates that the model achieves a perfect precision score of 1.00 at a confidence threshold of 0.929, while the F1–Confidence Curve shows an optimal F1-score of 0.73 at a confidence level of 0.39, demonstrating the best balance between precision and recall. Real-time testing further confirms that the system can detect papules with high confidence (88%), but confidence levels for comedones (31%) and nodules (29%) remain low due to visual similarity and non-ideal lighting conditions. Overall, the results indicate that the YOLOv8-based system is capable of performing real-time acne detection with acceptable accuracy. However, further improvements in dataset diversity and annotation quality are required to enhance performance, particularly for comedone detection.
A RAG-Based Academic Information Chatbot Using Lightweight LLaMA and Indo-Sencence-BERT Saman, Muhamad; Alfarisy, Gusti Ahmad Fanshuri; Amelia, Rizky; Fadhliana, Nisa Rizqiza
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i3.38150

Abstract

In the current digital era, Institut Teknologi Kalimantan (ITK) encounters challenges in delivering academic information that is fast, accurate, and easily accessible to students, lecturers, and academic staff. Access to important information such as administrative procedures, report writing guidelines, and academic policies remains largely reliant on manual systems and static handbooks. To address this issue, this study investigates a chatbot system utilizing the Retrieval-Augmented Generation (RAG) framework through LLaMA model. The chatbot combines semantic retrieval and natural language generation to provide relevant and accurate answers based on existing academic documents. Evaluation was conducted on two lightweight LlaMA models: 1.5 and 3B parameters. Furthermore, different embedding vector also evaluated along with Indo-Sentence-BERT as well as the chunking size. The most optimal configuration was achieved using LLaMA 3B as the generative model and Indo-Sentence-BERT as the retriever, with a chunk size of 200 tokens and an overlap of 10 tokens. This setup achieved a RAGAS score of approximately 0.9, a competitive MRR of 0.5, and response latency under 1 second. Although LLaMA 1B recorded a higher MRR (0.6), its low RAGAS score made it less favorable. Overall, the LLaMA 3B and Indo-Sentence-BERT configuration is recommended to enhance the efficiency of academic information retrieval at ITK.
Analysis of Spotify User Sentiment to Improve Customer Satisfaction Using Opinion Mining and Latent Dirichlet Allocation Based on E-Satisfaction Dimensions Samjas, Mutawakkil; Darmawan, Armin
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i3.38480

Abstract

This study aims to enhance Spotify customer satisfaction by analyzing user reviews on the Google Play Store using sentiment analysis techniques and identifying relevant topics related to customer satisfaction based on the dimensions of electronic satisfaction. The methods used in this analysis are Support Vector Machine (SVM), Naïve Bayes (NB), and Latent Dirichlet Allocation (LDA). The results show that SVM is the most effective technique for text classification, with accuracies of 87%, 87%, 81%, and 84%, respectively, along with precision, recall, and F1-score of 0.93, 0.93, and 0.84. LDA was utilized to extract various topics within the e-satisfaction dimensions, with serviceability emerging as the top priority for improvement. Identified topics include connectivity and accessibility, performance and user experience, premium services, app quality, content and playlists, app features, and sound/music quality. These findings suggest that improvements in server infrastructure, the implementation of AI-driven chat support, enhanced ad management, and improved song lyrics databases could substantially enhance Spotify's customer satisfaction.
Analysis of SQL Injection and Cross-Site Scripting (XSS) Attacks on Web Server Logs Using Machine Learning Septian, Adi; Rahman, Atep Aulia
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i3.38397

Abstract

The increasing complexity of cyber threats requires accurate detection systems to identify attack patterns on web servers. This study aims to detect SQL Injection and Cross-Site Scripting (XSS) attacks in Nginx access logs using machine learning algorithms. Log data were processed through regular expressions for parsing and labeling, resulting in 1,650,615 samples. Data imbalance was addressed using a combination of ADASYN and Random Undersampling. Two algorithms, Random Forest and Support Vector Machine (SVM), were compared based on accuracy, precision, recall, F1-score, and ROC curve metrics. The results show that Random Forest achieved the best performance with 99.92% accuracy, 99.94% F1-score, and 0.9994 AUC, while SVM obtained an accuracy of 96.45%. The combination of resampling and ensemble learning significantly enhances the effectiveness of log-based attack detection, providing a promising foundation for the development of adaptive Intrusion Detection Systems (IDS) in web server environments.
Robustness Testing of TrOCR for Multi-Condition Food Ingredient Labels Detected By YOLO Chairunnisa, Charina Mutiara; Husni, Nyayu Latifah; Kusumanto, RD.
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 2 (2025): July 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i3.37301

Abstract

This study aimed to develop an automatic text extraction system for ingredient labels by integrating YOLOv8 for object detection and a Transformer-based Optical Character Recognition (OCR) for text recognition. YOLOv8 was trained to detect and crop the label area in the image, while TrOCR was used to extract text from the cropped bounding box. The evaluation involved 16 sample image inputs under various conditions, including background color (Monochrome and RGB), languages (Bahasa Indonesia and English), and text formatting (single-line and multi-line). The results indicated that TrOCR performed well in single-line format, but struggled with multi-line format and longer text, even omitting words. Character and word error rates reached up to 100% for this complex layout. 
Intelligent Alert System With Yolo V8 Algorithm for Early Detection of Microsleep In Vehicle Drivers Putri, Ria Citra Desiany; Kusumanto, Kusumanto; Sari, Dewi Permata
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i3.37766

Abstract

Microsleep is a brief state of sleep that occurs suddenly without the person being aware of it and poses a serious risk to drivers, especially on long journeys. This study developed an intelligent alert system based on the YOLOv8 algorithm for the early detection of microsleep in drivers in real time by analyzing the state of the eyes and the position of the head. Using 3,458 annotated facial images as training data, the model was implemented on the Raspberry Pi platform for local processing without cloud dependency. The system activates a buzzer and warning light when it detects signs of drowsiness. Test results show the effectiveness of this method in the early detection of microsleep with 90.3% precision, 91.3% accuracy, 96.8% recall, and an F1 score of 93.9%. It has been shown to function optimally in a variety of lighting conditions to improve road safety.
Hybrid Reinforcement and Evolutionary Learning Model for Adaptive Pathway Optimization In Computer Networks Education Anggraeni, Sherly Rosa; Wahyudi, Dian Julianto; Silviariza, Waode Yunia; Ro’is, Rachmy Rosyida; Ranggianto, Narandha Arya
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i3.38398

Abstract

This paper introduces a Hybrid Reinforcement and Evolutionary Learning Model developed to optimize adaptive learning pathways in computer network education. Traditional uniform curricula often struggle to accommodate diverse learner profiles, resulting in knowledge gaps across hierarchical concepts such as OSI layers, routing protocols, and security mechanisms. The proposed model integrates Deep Knowledge Tracing (DKT) with Long Short Term Memory (LSTM) networks for real-time estimation of learners’ knowledge states, Proximal Policy Optimization (PPO) for dynamic sequential content selection, and a Genetic Algorithm Particle Swarm Optimization (GA–PSO) hybrid for global pathway refinement under constraints such as prerequisites and time limits. The model was evaluated using real learner data from an e-learning platform and achieved an average final mastery score of 0.867, quiz accuracy of 0.822, and an F1-score of 0.880 for path recommendations outperforming baseline models such as static curricula (0.740 mastery) and DKT+PPO (0.824 mastery) by 5–17%. Ablation studies validated the synergistic contribution of each component, with the GA–PSO module enhancing optimization efficiency by approximately 10%. Overall, these findings demonstrate that the proposed model offers superior personalization, learning efficiency, and adaptability, marking a significant advancement in AI-driven education for computer networks.
Facial Expression Detection of Autism Children Using ResNet-50 in Convolutional Neural Network Algorithm Prihatini, Ekawati; Muslimin, Selamat; Darmawan, Muhammad Rizki
Indonesian Journal of Artificial Intelligence and Data Mining Vol 8, No 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v8i3.37755

Abstract

Facial expression detection in children with autism presents unique challenges due to limitations in verbal communication and social responses. This study develops a Convolutional Neural Network (CNN) model using the ResNet-50 architecture to improve the recognition accuracy of five expression categories: angry, fear, sad, neutral, and happy. A dataset of 3,030 images was divided into training and testing sets (60:40), with data augmentation and hyperparameter tuning applied using the Adam optimizer. The model achieved 89% validation accuracy and 84.49% testing accuracy, along with 86.78% precision and 80.69% recall. Evaluation on 25 test images showed an 84% success rate. These results indicate that ResNet‑50 effectively extracts facial features and classifies expressions with high accuracy, demonstrating potential as a communication aid in autism therapy. Future improvements include adding more diverse training data and optimizing model parameters.
Identification of Mental Health for Generation Z Using Machine Learning Algorithm Retnowati, Sri; Aisyah, Siti
Indonesian Journal of Artificial Intelligence and Data Mining Vol 9, No 1 (2026): March 2026
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v9i1.38528

Abstract

Mental health issues such as stress, anxiety, and trauma have become significant challenges, particularly among Generation Z. The lack of effective early detection tools has hindered efforts to address these problems promptly and accurately. This study aims to develop a machine learning-based classification model to detect potential mental health conditions using standardized psychological instruments: DASS-21, STAI, and ACE. Data were collected from 733 youths aged 17–24, of whom 212 exhibited signs of risk. After cleaning and preprocessing, 58 features were retained from the initial 92. Several machine learning models such as Logistic Regression, Support Vector Machine (SVM), and Random Forest were evaluated using class balancing techniques including SMOTE and class weighting. Evaluation metrics are included accuracy, recall, precision, F1-score, and ROC AUC. Logistic regression achieved the highest performance, with 94% accuracy, 100% recall, 82% precision, and an F1-score of 0.90. The ROC AUC reached 99.5%, indicating excellent discriminative ability. This research highlights the effectiveness of machine learning for early detection of mental health conditions and supports its integration into scalable, technology-based mental health screening tools, particularly for at-risk youth populations.