Journal of Intelligent Decision Support System (IDSS)

Butterfly species identification using glcm features and edge detection using KNN (K-Nearest Neighbor) and decision tree algorithm (C.45)

Muhamad Hasan — 2026-03-30

Butterflies are insects come from the kingdom Animalia, which are the Insecta class, the Lepidoptera order, and the sub-order of Rhopalocera. Butterflies can classified according to the patterns found on the butterfly's wings. Butterfly species have different patterns based on pigment, scale structure, and sunlight fall structure. The weakness of the human eye in specific the patterns in butterflies is the foundation in basis butterfly identification based on pattern recognition. This study used 3 butterfly species: Adonis, Black Hairstreak, and Gray Hairstreak. The butterfly dataset used was 150 which were obtained online. The pre-processing stage used segmentation and edge detection methods. The feature extraction stage used the Gray-level Co-occurrence Matrix (GLCM) method which extracted 8 shape and texture features including area, perimeter, metric, eccentricity, contrast, correlation, energy, and homogeneity. Classification phase used K-Nearest Neighbor (KNN) method with the values of k = 3, 5, 7, 9, 11, 13, 15, 17, and 19 as well as the Decision Tree method (C.45). The results of the identification of butterflies with the highest accuracy were obtained by the KNN Algorithm on the testing with a value of k = 3 of 93.33%, and the accuracy results using the Decision Tree method (C.45) is 84.44% while the results of identification using an application made using the GUI Matlab2017 with the KNN algorithm obtained an accuracy of 93.33% with a value of k= 3.

A mixed integer linear programming approach for last-mile e-commerce optimization through micro-fulfillment centers

Fristi Riandari — 2026-03-30

The rapid growth of e-commerce increases the complexity of last-mile delivery due to high distribution costs, urban congestion, and increasingly tight delivery time demands. This study proposes a Mixed Integer Linear Programming (MILP) approach to optimize e-commerce last-mile distribution through the determination of Micro-Fulfillment Centers (MFCs). The model simultaneously determines (i) the locations of candidate MFCs to be opened and (ii) the allocation of demand zones to selected facilities, with the objective of minimizing the total network cost consisting of fixed facility costs and variable last-mile service costs. Service quality is enforced through a hard service level agreement (SLA) mechanism by limiting allocation to only pairs of facility zones that meet a certain travel time threshold, while operational feasibility is guaranteed through capacity constraints at each MFC. The model outputs are implementable in the form of selected MFC locations, zone allocation maps, and performance indicators for evaluation, including total cost decomposition, weighted travel time metrics, and facility capacity utilization to identify potential bottlenecks. Numerical illustrations show that the MILP formulation yields feasible location–allocation decisions with respect to SLA and capacity, while avoiding the “closest/fastest” heuristic that can potentially lead to facility overload. This framework supports decision-makers in designing efficient, responsive, and scalable last-mile networks, and can be extended to incorporate demand uncertainty, SLA penalties (soft-SLAs), multi-echelon structures, and sustainability objectives.

Disinformation propagation modeling in digital information warfare using hybrid GNN and LSTM

Jonson Manurung — 2026-03-31

The rapid growth of digital information warfare has enabled the widespread dissemination of disinformation, posing serious challenges for detection systems. However, most existing approaches treat disinformation detection as a static classification problem and fail to consider the network structure and temporal dynamics of information spread. This study proposes a hybrid deep learning model that combines Graph Attention Networks (GAT) and Bidirectional Long Short-Term Memory (BiLSTM) with a cross-attention mechanism to capture both structural and temporal patterns of disinformation propagation. The proposed model was evaluated using three datasets: the PHEME rumor dataset, a large-scale Twitter and X crisis dataset, and a synthetically generated defense simulation dataset. Experimental results show that the model achieves strong performance, with 92.47% accuracy in classification, 89.63% precision in cascade prediction, 87.91% F1-score in source identification, and a mean absolute error of 0.183 in predicting spread dynamics, outperforming several baseline methods. These findings demonstrate that integrating network-based and temporal modeling can significantly improve disinformation detection performance. Future research will focus on incorporating multimodal data, real-time processing, and cross-platform learning to enhance the robustness of the proposed approach.

Implementation of agile SCRUM and Agile UX in the design of the Pelabuhan Ratu application "SI-KIKAN"

Uus Firdaus — 2026-03-30

This study addresses the inefficiency and uncertainty in accessing real-time fish stock information at the Pelabuhan Ratu Fish Market, where traditional manual communication methods remain dominant and hinder effective transactions. The objective of this research is to develop a user-centered mobile application, SI-KIKAN, that provides real-time fish availability information while ensuring high usability for users with diverse levels of digital literacy. This study adopts a design and development approach by integrating Agile Scrum and Agile User Experience (UX) methodologies, enabling iterative system development, continuous user feedback, and adaptive design refinement. Data were collected through field observations, stakeholder interviews, and usability testing involving 30 respondents, and analyzed using qualitative techniques and the System Usability Scale (SUS). The results indicate that the SI-KIKAN application successfully improves information transparency and operational efficiency, as evidenced by a high SUS score of 86.17, categorized as “Excellent,” demonstrating strong usability, learnability, and user satisfaction. The implementation of Agile Scrum facilitated efficient management of development complexity, while Agile UX ensured that the interface design aligned with user needs and field conditions. This study implies that the integration of Agile Scrum and Agile UX is an effective approach for developing user-centered digital solutions in traditional sectors, particularly in improving information accessibility and supporting digital transformation in fisheries markets.

Distributed cyber defense framework based on federated learning for attack detection in defense infrastructure

Hondor Saragih — 2026-03-31

Cyber threats targeting defense infrastructure have escalated in complexity, rendering centralized intrusion detection systems insufficient due to their inability to guarantee data privacy across distributed military nodes. This study proposes a distributed cyber defense framework that employs federated learning to enable collaborative model training without transmitting raw network traffic beyond individual nodes. The framework integrates an adaptive aggregation strategy combining FedAvg and FedProx, a hybrid deep learning architecture consisting of convolutional neural networks and long short term memory networks, an autoencoder module for unsupervised anomaly detection, a Byzantine robust aggregation mechanism, and post hoc explainability through SHAP and LIME. Experiments were conducted on CIC IDS 2017, CIC IDS 2018, UNSW NB15, and a synthetically generated military network traffic dataset. The proposed framework attained a peak accuracy of 98.74% and an F1 score of 98.12% on CIC IDS 2017, consistently outperforming five baseline methods by up to 5.29 percentage points in F1 score. Future work will investigate differential privacy integration and model compression for deployment on resource constrained tactical edge devices.

Identifying key patterns of college student’s background through exploratory data analysis

Sitti Rahmah Jabir — 2026-03-31

The declining of student interest had forced universities to examine the characteristics of each student. According to higher education statistics on the number of new students, fluctuating values have been found in recent years. Several research used exploratory data analysis (EDA) approach to analyze new student admissions data. EDA is offered a summary of the dataset analysis and preliminary findings. There are variables decided to be dropped because consisted high number of missing values. On the other hand, some data filled with mean and mode because the number of missing not more than 20%. The missing values in each of attribute might be cleaned using another way. The admission team in university might encourage the registrants to complete and input correct data to the system. Based on the visualization, we found that some college students applied to university from several background of area, demographic and etc. The marketing division might apply another strategy is area had small number of college which is Kalimantan. Public health, computer science and insutry technology are major that have potential to be promoted due to the job prospects.