cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
Akim Manaor Hara Pardede
Contact Email
jaiea@ioinformatic.org
Phone
+6281370747777
Journal Mail Official
jaiea@ioinformatic.org
Editorial Address
Jl. Gunung Sinabung Perum. Grand Marcapada Indah. Blok. F1. Kota Binjai. Sumatera Utara
Location
Unknown,
Unknown
INDONESIA
Journal of Artificial Intelligence and Engineering Applications (JAIEA)
Published by Yayasan Kita Menulis
ISSN : -     EISSN : 28084519     DOI : https://doi.org/10.53842/jaiea.v1i1
Core Subject : Science, Engineering,
Automotive Engineering Computer Science & IT Control & Systems Engineering
The Journal of Artificial Intelligence and Engineering Applications (JAIEA) is a peer-reviewed journal. The JAIEA welcomes papers on broad aspects of Artificial Intelligence and Engineering which is an always hot topic to study, but not limited to, cognition and AI applications, engineering applications, mechatronic engineering, medical engineering, chemical engineering, civil engineering, industrial engineering, energy engineering, manufacturing engineering, mechanical engineering, applied sciences, AI and Human Sciences, AI and education, AI and robotics, automated reasoning and inference, case-based reasoning, computer vision, constraint processing, heuristic search, machine learning, multi-agent systems, and natural language processing. Publications in this journal produce reports that can solve problems based on intelligence, which can be proven to be more effective.
Arjuna Subject : Ilmu Komputer (semua kategori) - Kecerdasan Buatan
Articles 430 Documents
 35   36   37   38   39 
Application of Decision Tree Algorithms to Classify the Sales Results of Kangen Kripik Sme Products Adila G Khiqmatiar Muchsin; Nining Rahaningsih; Irfan Ali; Dadang Sudrajat; Saeful Anwar
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1854

Abstract

Micro, Small, and Medium Enterprises (MSMEs) play a vital role in strengthening the national economy; however, many still face challenges in managing and analyzing sales data effectively. This study aims to classify product sales results at UMKM Kangen Kripik Mang Acep by applying the Decision Tree algorithm as a data classification method based on machine learning. A quantitative experimental approach was employed to evaluate the model’s performance using one-year sales data, including attributes such as product variants, sales volume, sales channels, and marketing regions. Data processing was conducted using RapidMiner software following the Knowledge Discovery in Databases (KDD) framework, which includes data selection, preprocessing, transformation, data mining, and model evaluation. The results indicate that the Decision Tree algorithm successfully classified sales regions (Garut, Bandung, and Sumedang) with an accuracy rate of 96.48%, identifying “Units Sold (pcs)” as the most influential attribute for distinguishing marketing areas. These findings demonstrate that the Decision Tree method is not only effective in improving data analysis efficiency but also provides valuable strategic insights for data-driven business decision-making in MSMEs
Literature Review: Transitioning usage from BFS and DFS to Heuristic Search in the Modern AI Era Syahputra, Fahmy; Sabrina, Elsa; Sahendra Chan, M Fajar; Fali, Rifki; Fattah, Muhammad; Hendratmo, Joko; Ardiansyah, Fadhil
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1856

Abstract

Uninformed search algorithms, specifically Breadth-First Search (BFS) and Depth-First Search (DFS), encounter significant scalability limitations when addressing complex problem spaces in modern Artificial Intelligence (AI) ecosystems. This study investigates the paradigm shift toward intelligent heuristic algorithms through a systematic literature review and comparative analysis of 24 recent academic sources. The evaluation focuses on three primary domains: logical problem solving, robotic navigation, and data infrastructure management. Results demonstrate that heuristic methods, such as A-Star and hybrid variants like PrunedBFS, offer superior time efficiency and memory optimization for autonomous navigation and massive computing tasks. Nevertheless, classic algorithms retain functional relevance for specific scenarios requiring exhaustive exploration. Furthermore, this study reveals that algorithmic evolution has fundamentally transformed digital infrastructure, driving a shift from Search Engine Optimization (SEO) to Answer Engine Optimization (AEO) and necessitating adaptive cybersecurity architectures. The research concludes that the future of AI development relies not on substitution, but on a collaborative synthesis integrating the robustness of classic methods with the adaptability of modern heuristics.
Comparison of Memory Efficiency and Computation Time of Bubble Sort, Insertion Sort, and Intro Sort Algorithms Using the C++ Programming Language Nur Fatmaluna, Shabina; Nazwa, Nazwa Gista Aulia; Adhiel, Adhiel Rahma; Salma, Salma Aulia; Imam, Imam Prayogo Pujiono
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1858

Abstract

Data sorting is a fundamental step in the computer process that greatly affects the effectiveness of programs and overall system performance. In this study, three sorting algorithms, namely Bubble Sort, Insertion Sort, and Intro Sort, are analyzed and compared using recursive and iterative approaches. Bubble Sort serves as a basic algorithm example to understand the basic idea of element exchange, while Insertion Sort was chosen for its efficiency on small and nearly sorted datasets. Intro Sort, as a combination algorithm that integrates Quick Sort, Heap Sort, and Insertion Sort, was studied to reveal how its adaptive mechanism can provide more optimal results. The testing was conducted by measuring execution speed, sorting stability, and memory usage efficiency. The findings from this study show that Bubble Sort ranks lowest in terms of performance and is less suitable for large data sets. Insertion Sort shows better results on small data sets and those with similar patterns. Intro Sort emerges as the most effective algorithm with stable processing time, high adaptability, and faster and more efficient sorting results for various data sizes. Overall, this study emphasizes the importance of choosing a sorting algorithm that suits the characteristics of the data and the needs of the application. The combination of adaptive strategies such as those in Intro Sort is the best solution for current data processing, which demands high speed and efficiency.
Comparison of Balancing Strategies for Classifying Guava Fruit Diseases Putri Nabilla; Suarna, Nana; Bahtiar, Agus; Rahaningsih, Nining; Prihartono, Willy
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1859

Abstract

The problem of class imbalance often poses an obstacle in deep learning-based image classification, especially in the domain of digital agriculture. The imbalance in data distribution makes it easier for models to recognize the majority class, while performance for the minority class declines. This study aims to analyze the effectiveness of three strategies for handling class imbalance: Weighted Loss Function, Oversampling, and a combination of Weighted Loss and Oversampling, in improving the performance of image classification of guava fruit diseases using a transfer learning-based MobileNetV2 architecture. The dataset consists of 3,784 images of three disease classes, namely Anthracnose, Fruit_Fly, and Healthy_guava, which show an imbalanced distribution. The research was conducted through the stages of Exploratory Data Analysis (EDA), pre-processing, augmentation, model training with four scenarios, and evaluation using Accuracy, Precision, Recall, F1-Score, and Macro Average F1-Score. The results showed that the Combination model (Oversampling and Weighted Loss) performed best on the minority class with an F1-score of 0.9630, the highest among all models. The Oversampling strategy produced the highest Macro F1-score of 0.9617, while Weighted Loss provided a significant improvement in classification sensitivity but was still below the combination model. Thus, it can be concluded that the combination strategy is the most effective approach in improving the sensitivity of the model to minority classes, while Oversampling excels in the overall performance stability of the model.
Predicting Student Academic Performance Based on Learning Habits Using XGBoost and SHAP Latifah, Siti; Martanto; Dana, Raditya Danar; Dikananda, Fatihanursari; Hayati, Umi
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1860

Abstract

This study developed a model for predicting student academic achievement based on learning habits using the XGBoost algorithm and SHAP interpretability techniques. The secondary dataset contains 1,000 entries and 16 variables (for example, hours of study per day, mental health, frequency of exercise, social media use, hours of sleep) pre-processed including cleaning, imputation, encoding, and normalization before being divided into train–test (80:20) and validated using 5-fold CV. Three models were tested: Linear Regression, Random Forest, and XGBoost. Evaluation using RMSE, MAE, and R² showed that XGBoost achieved RMSE = 0.335, MAE = 0.266, and R² = 0.882, while Linear Regression showed the best performance according to R² in certain configurations (R² = 0.888; RMSE = 0.326). SHAP analysis revealed that the most influential features were hours of study per day, mental health scores, exercise frequency, duration of social media use, and hours spent watching Netflix. The findings confirm that students' study habits and psychological conditions are the main determinants of academic achievement variation; the use of interpretable features strengthens the readability of the model for education stakeholders. Research recommendations include testing the model on longitudinal datasets, integrating socioeconomic factors, and implementing data privacy procedures before institutional-scale implementation.
Optimization of Classification of Tea Leaf Disease Images Using LBP–HOG and MobileNetV2 Ezar Qotrunnada; Nurdiawan, Odi; Dikananda, Arif Rinaldi; Putra, Aris Pratama; Nurhakim, Bani
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1861

Abstract

This study was motivated by the need for an accurate and efficient system for detecting tea leaf diseases, given that the current method Manual identification has limitations in terms of consistency, speed, and It also depends on expert labor. To address these challenges, the study It developed a classification model for detecting diseases in tea leaves using a combination of features Local Binary Patterns (LBP) and Histogram of Oriented Gradients (HOG) integrated with the MobileNetV2 architecture. The research method includes the following stages: importing the dataset, data partitioning, exploratory data analysis (EDA), preprocessing, features, and training four model scenarios: baseline MobileNetV2, LBP-based model, HOG-based model, and hybrid LBP–HOG model. Evaluation is done with the metrics of accuracy, precision, recall, and F1-score. The results show that the baseline model achieved 91.67% accuracy, the LBP model achieved 60.67%, the HOG model achieved 68.67% accuracy, and the hybrid model achieved 66.67% accuracy. These findings indicate that MobileNetV2 is still the most optimal model, but the integration of texture features and gradients provides a deeper understanding of the characteristics of disease patterns. This study emphasizes the importance of exploring classic features to enriching visual representation in lightweight CNN models, as well as providing a contribution to the development of plant disease diagnosis systems that are efficient.
Mitigating Imbalanced Citrus Disease Image Datasets with Oversampling Gunawan, Arya; Suarna, Nana; Bahtiar, Agus; Marthanu, Indra Wiguna; Kaslani
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1862

Abstract

Dataset imbalance is a critical challenge in plant disease image classification because it causes bias towards the majority class. This study evaluates the effectiveness of augmentation-based oversampling techniques on the classification performance of citrus leaf images using the MobileNetV2 architecture. The four leaf disease classes classified include Greening, Fresh, Canker, and Blackspot. The dataset was obtained from a public repository and processed through preprocessing (resize, normalization) and augmentation (rotation, flipping, zoom) stages. The model was trained and tested in two scenarios: baseline (unbalanced data) and mitigation (data balanced through augmentation). The experimental results show that the mitigation approach was able to increase accuracy from 91.92% to 93.94%. The F1-score, precision, and recall values also increased significantly, especially in the minority class. Evaluation using a confusion matrix reinforced the finding that augmentation-based oversampling is effective in reducing classification errors. This study shows that the integration of augmentation techniques and MobileNetV2-based transfer learning can significantly improve classification performance and contribute to the development of early detection systems for plant diseases in precision agriculture.
Segmentation of Coffee Purchasing Behavior Based on Transaction Time Using the K-Means Algorithm Yuslia Devitri; Rahaningsih, Nining; Ali, Irfan; Prihartono, Willy
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1863

Abstract

This studyaims to identify customer behavior patterns based on the time of purchaseof beverages at a coffee shop using the K-Means method.Transaction data includes purchase time, payment type, product name,time category, day, and month. The research stages include data cleaning, time attribute transformation, and numerical feature normalization. The optimal number of clustersis determined through testing k = 2–10 with four evaluation metrics,namely Inertia, Silhouette Score, Davies–Bouldin Index, and Calinski–HarabaszIndex. Based on the validation results, k = 3 was selected because it provided the best balancebetween compactness and cluster separation. The clustering results showedthree main customer groups based on transaction time trends:nighttime buyers with a peak around 10:27 p.m., afternoon to early evening buyerswith a centroid of 7:01 p.m., and morning to noon buyers with a centroid11:13. The frequency distribution indicates that the morning–afternoon buyer groupis the largest, while the early evening–night group is thesmallest. Visualization of scatter plots, boxplots, and time category graphsemphasizes the differences in characteristics between clusters. Overall,this study proves that K-Means is effective in mapping the temporal patternsof customer behavior. These findings can be used to develop time-based marketing strategies, operational arrangements, and product stock management,as well as form the basis for further analysis in the industry.
Comparative Analysis of Durian Leaf Disease Classification Using Transfer Learning VGG16, InceptionV3, and U-Net Nafisa Maysa Salma; Kurniawan, Rudi; Nurhakim, Bani; Bahtiar, Agus; Narasati, Riri
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1864

Abstract

Image-based durian leaf disease detection presents challenges due to high visual similarity among symptoms and the limited, imbalanced dataset. This study compares three deep learning architectures VGG16, InceptionV3, and U-Net encoder-based—using transfer learning for classifying five durian leaf conditions. The dataset of 4,437 images underwent preprocessing, augmentation, and preliminary segmentation using U-Net to enhance focus on leaf regions. Fine-tuning was applied to the upper layers of each model to adapt feature representations to tropical leaf characteristics. The results indicate that InceptionV3 achieved the most stable and accurate performance with an accuracy of approximately 0.66, while VGG16 showed balanced results but was more prone to overfitting. U-Net proved effective for segmentation but less optimal as a classifier due to loss of small-scale lesion details. Overall, the findings demonstrate that combining U-Net segmentation with CNN-based transfer learning improves disease identification performance, particularly under limited data conditions.
Comparison of Graph-Based Filtering and Non-Local Means Techniques in Diabetic Retinopathy Classification Gita Antar Wulan; Irawan, Bambang; Faqih, Ahmad; Putra, Aris Pratama; Nurhakim, Bani
Journal of Artificial Intelligence and Engineering Applications (JAIEA) Vol. 5 No. 2 (2026): February 2026
Publisher : Yayasan Kita Menulis

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59934/jaiea.v5i2.1869

Abstract

Classification of diabetic retinopathy (DR) based on retinal images is important for early detection, but is often hampered by poor image quality such as noise, uneven lighting, and low contrast. This study analyzes the effect of applying three image filtering techniques, namely Graph Laplacian Filtering (GLF), Graph Convolutional Network (GCN), and Non-Local Means (NLM), on improving the performance of Diabetic Retinopathy classification. The three methods were compared with a baseline model without filtering using VGG16 and evaluated through accuracy, AUC, loss, and image quality metrics such as PSNR, SSIM, MSE, and RMSE.The results showed that graphical and spatial filtering did not always improve classification performance, as VGG16 Fine-Tuning without filtering achieved the highest accuracy of 97.84%. Combinations with NLM, GCN, and Graph Laplacian resulted in lower accuracy due to the smoothing effect that removed important microfeatures on the retina. However, NLM remained effective in reducing noise without disturbing edge structures. These findings confirm that improving image visual quality does not always correlate with CNN accuracy, so preprocessing must focus on preserving diagnostic features.  

Page 37 of 43 | Total Record : 430

 35   36   37   38   39 

Filter by Year

2021 2026


Reset
Filter By Issues
All Issue Vol. 5 No. 2 (2026): February 2026 Vol. 5 No. 1 (2025): October 2025 Vol. 4 No. 3 (2025): June 2025 Vol. 4 No. 2 (2025): February 2025 Vol. 4 No. 1 (2024): October 2024 Vol. 3 No. 3 (2024): June 2024 Vol. 3 No. 2 (2024): February 2024 Vol. 3 No. 1 (2023): October 2023 Vol. 2 No. 3 (2023): June 2023 Vol. 2 No. 2 (2023): February 2023 Vol. 2 No. 1 (2022): October 2022 Vol. 1 No. 3 (2022): June 2022 Vol. 1 No. 2 (2022): February 2022 Vol. 1 No. 1 (2021): October 2021 More Issue