Jurnal Teknik Informatika dan Sistem Informasi

The Development of Web-Based Geographic Information System for University Facilities

Satria Berliano Manzi — 2025-12-15

The University of Lampung is one of the state universities in Lampung which has an area of around 65 hectares, has various facilities spread throughout the campus area. The diversity of facilities in the campus environment often causes difficulties for students and visitors in finding and recognizing these facilities, and it is very difficult and limited to get access and the location of the facilities. This study aims to develop a website-based Geographic Information System (GIS) that can be a source of information about facilities at the University of Lampung, test system functionality, and measure effectiveness, efficiency, and user satisfaction. System development uses the Agile Scrum method with 5 sprints. The system was developed using Next.js as the main framework, Leaflet.js for map visualization, and Supabase as a backend service. System testing is carried out through three methods: blackbox testing with decision table techniques, usability testing with 20 respondents, and heuristic evaluation by three expert evaluators. The test results showed that the system successfully completed 8 test items with 27 test combinations, achieving an effectiveness level of 97.5%, efficiency of 96.67%, and a user satisfaction score of 1,725 on a scale of 7 on the Post-Study System Usability Questionnaire (PSSUQ). The heuristic evaluation identified 22 problems which were then resolved through 18 recommendations for improvement. This system successfully provided detailed information on campus facilities and their locations accurately and can be easily accessed via the website.

Centralized Network Management for Configuration and Recovery Automation using Paramiko and Django Framework

Hillman Akhyar Damanik — 2025-12-15

This study aims to implement network automation configuration, improve time efficiency, accelerate the centralized configuration and recovery process on MikroTik vCHR and RB70 series router devices using Paramiko, Django framework, and secure protocol SCP. The results of the experiments conducted showed a significant increase in time efficiency and a reduction in configuration and recovery time on 10 vCHR and RB750 router device nodes consisting of 496 script lines. Each router was allocated a data transfer speed of 1024 Kbps, for a backup file capacity range of 47 KB to 59 KB. The transfer speed has an estimated backup time calculated using a mathematical approach, which shows that each vCHR and RB750 device takes about 7.15 seconds to 7.79 seconds for the backup process. The uptime and downtime testing parameters including the percentage of ICMP loss, response time and unavailable ICMP), showed that during the configuration sending process, network performance remained optimal with an average availability reaching 100% and the results of the throughput required by the vCHR-C1 router were 50 Kbps, vCHR-C2 49 Kbps, vCHR-C3 49 Kbps, vCHR-C4 32 Kbps, vCHR-C5 50 Kbps and RB750-C6 38 Kbps, RB750-C7 37 Kbps, RB750-C8 37 Kbps, RB750-C9 38 Kbps, RB750-C10 22 Kbps. The entire configuration and recovery process went well without any data loss or errors in data backup

Support Vector Machine Algorithm Optimization for Sentiment Analysis using Bayesian Optimization

Muhammad Resa Arif Yudianto — 2025-12-15

This study examines the effect of Bayesian Optimization in improving the performance, computational efficiency, and sustainability of Aspect-Based Sentiment Analysis models using Support Vector Machine (SVM). A dataset consisting of 988 customer reviews about Borobudur Temple, classified into six dimensions: Attractiveness, Facilities, Accessibility, Visual Image, Price, and Human Resources is used to compare two scenarios, namely Baseline SVM and SVM enhanced with Bayesian Optimization (BO). Important metrics used include accuracy, computational duration, energy usage, and carbon emissions. The results show that BO significantly improves accuracy, especially on difficult aspects such as Facilities (from 0.7294 to 0.8682) and Price (from 0.8047 to 0.9576). The most complicated aspect, namely visual image due to the very minimal number of datasets (unbalanced), achieved an increase in accuracy from 0.6729 to 0.72. In addition, BO reduces training time, especially for resource-intensive tasks such as the visual image aspect, reducing training time from 13.04 seconds to 9.4 seconds. Substantial reductions in energy consumption and CO₂ emissions are seen in line with sustainable machine learning principles. The hyperparameter adaptability of SVM, with linear kernels performing well in simpler tasks, while polynomial and sigmoid kernels improve performance for more complex parts. BO substantially alleviates the limitations of Baseline SVM, offering a robust, efficient, and environmentally friendly solution for ABSA. Future research can explore more enhancements for complex tasks to improve performance and efficiency.

Application of Integrated Methodology in Academic Process Engineering at Universities

Tiur Gantini — 2025-12-15

Rapid technological developments in the era of globalization require universities to continuously improve their academic systems. When academic systems undergo renewal, related business processes must also be reengineered to align with ongoing activities. This study aims to analyze the effectiveness of integrating four methods—Value Stream Mapping (VSM), Voice of Customer (VOC), Voice of Business (VOB), and Pick Chart—in reengineering academic business processes at a university in Bandung. The research methodology consists of three phases: problem formulation, theoretical review, and business process reengineering design. Data was collected through interviews with stakeholders, including academic staff, lecturers, and students, as well as direct observation of 14 (fourteen) Standard Operating Procedures. The VSM method was used to map the current process and identify problems, VOC captured customer complaints, VOB measured institutional-level problems, and Pick Chart prioritized solutions based on difficulty and impact of implementation. The results of the study show that this integrated framework effectively identifies inefficiencies and proposes targeted improvements. For the grade revision process, the re-engineered business process reduced processing time from approximately one week to 55 (fifty-five) minutes, while eliminating paper waste through the implementation of digital forms. This study concludes that the integration of VSM, VOC, VOB, and Pick Chart provides a comprehensive and systematic framework for re-engineering academic business processes that can be replicated by other higher education institutions.

Implementation of Certainty Factor in Expert System for Oil Palm Disease Diagnosis

Nadiva Azro Fathinah — 2025-12-15

Diseases in palm oil plants are one of the main causes of palm oil production not being maximized, and can even result in crop failure. Farmers need to know the symptoms that occur in oil palm plants in order to diagnose and overcome the diseases that infect the palm oil plants. A system for early detection of disease in palm oil plants is needed in order to prevent a decrease in productivity. An approach that can be used for early diagnosis is an expert system. Expert systems not only provide a diagnosis, but also offer an explanation of the type of disease as well as practical and accurate treatment recommendations. This research applies one of the methods of the certainty factor method to an expert system that combines several symptoms to determine how likely a diagnosis is. This expert system involves 22 symptoms to diagnose six diseases in palm oil plants. The accuracy rate obtained from the application of the expert system with the certainty factor method in diagnosing diseases of oil palm plants based on data from five users shows a result of 100%. This shows that the expert system with the certainty factor method is accurate and can be applied to early detection of diseases that attack palm oil plants.

Interface Redesign Using Design Thinking, Sentiment, Topic Modeling, and Visual Evaluation

Rosa Salsa Saida — 2025-12-15

The digitalization era encourages the presence of various application-based services, including online shopping services. One such platform is FamiApps by FamilyMartID, developed as a convenient tool for ordering Family Mart Indonesia products. However, a low rating of 2.29 based on the latest data from Google Play Store indicates user dissatisfaction, which falls to capture the full scope of user concerns. This study aims to redesign the user interface (UI) and user experience (UX) of FamiApps by implementing Design Thinking, focusing on user needs. In the Empathize stage, sentiment analysis was conducted using the VADER method on 266 user reviews to identify negative sentiments, which were then further analyzed using Latent Dirichlet Allocation (LDA) to identify pain points. These findings served as the poundation for the Define stage to formulate user needs, followed by the Ideate and Prototype stages, incorporating Nielsen’s heuristics and Gestalt principles as interface design guidelines. The final design was evaluated using the System Usability Scale (SUS) with 35 respondents, resulting in a score of 82, categorized as “Excellent” in adjective ratings, “Acceptable” in acceptable ratings, and graded “A” in the grade scale. The findings demonstrate that a review-driven design approach can produce UI/UX solutions that are more relevant, adaptive, and contribute to enhancing user experience quality.

A Novel Minimax Regularization Framework for Enhancing Neural Network Robustness

Jincheng Zhang — 2025-12-15

In the development of deep learning, regularization techniques have been widely used to improve the generalization ability and robustness of models. However, traditional regularization methods are often based on a priori assumptions and fail to fully consider the performance of the model in the worst case. This paper proposes a regularization mechanism based on the Minimax theorem, attempting to introduce the idea of "worst-case adversarial" during the training process to improve the robustness of the model. Through experimental verification of the CIFAR-10 dataset, we observed that this method is slightly better than the standard multi-layer perceptron (MLP) model in multiple evaluation indicators and shows good generalization performance. This method has a wide range of applicability and can be extended to a variety of architectures including convolutional neural networks, graph neural networks, and natural language processing models.

Implementation of Retrieval Augmented Generation in a Web-Based Dermatological Chatbot System

Ivana Lucia Kharisma — 2025-12-15

Indonesia’s tropical climate, poor sanitation, and limited access to medical services especially in remote areas are key factors contributing to the high prevalence of skin diseases. Direct access to dermatologists remains difficult for many people. This study aims to develop a dermatological consultation Chatbot using a Retrieval Augmented Generation (RAG) approach, leveraging the LangChain framework, the LLaMA model, and the Qdrant vector database. The dataset includes 30 types of skin diseases sourced from the National Library of Medicine. The preprocessing stage involved whitespace normalization, removal of special characters, and handling of missing values to ensure data consistency before vectorization. Evaluation results showed high scores for Faithfulness (0.9429) and LLMContextRecall (0.9600), indicating that the responses were relevant and aligned with the source documents. However, a relatively low Precision score (0.4720) suggests a need for improved information accuracy. The Chatbot is integrated with the Chainlit platform, offering an interactive user interface that supports login, conversation history, and user feedback features to facilitate system development based on user input. The system demonstrated fast retrieval times (0.08–0.29 seconds), though answer generation remains slow due to CPU infrastructure limitations (255–283 seconds). Future improvements should focus on enhancing answer accuracy, optimizing the model's performance, enriching the medical reference dataset, and adding automated medical validation features to ensure the reliability of consultations. Therefore, this Chatbot system is expected to serve as a cost-effective and efficient alternative for providing initial information on skin conditions to individuals with limited access to healthcare services.

Effectiveness Analysis of Multimodal Feature Fusion in Herbal Leaf Classification

Riki Riyandi — 2025-12-15

This study aims to evaluate the performance of leaf image classification models based on feature fusion strategies that integrate shape, texture, and semantic representations. Three feature extraction methods were employed: Histogram of Oriented Gradients (HOG) for shape, Gabor Filter for texture, and Convolutional Neural Network (CNN) using MobileNetV2 for semantic features. Each feature type was tested using three classification algorithms: Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Random Forest (RF). Experimental results show that CNN features consistently outperformed the others, achieving the highest accuracy and F1-score, with a peak accuracy of 91.0% using CNN+SVM. In contrast, HOG and Gabor features resulted in significantly lower performance. Feature fusion—such as HOG+CNN and HOG+Gabor+CNN—did not improve performance and instead caused a notable decline, primarily due to the high dimensionality of HOG features, leading to the curse of dimensionality. Confusion matrix and ROC curve analyses confirmed that the CNN-based model achieved high inter-class separability, while models with fused features produced near-random predictions in several classes. These findings suggest that feature fusion does not inherently lead to better classification performance, particularly when dimensional imbalance is not addressed. The study recommends the use of single semantic features extracted from CNN for efficient and accurate leaf image classification, while also encouraging future research into adaptive fusion strategies such as feature weighting or multimodal integration.

Online Review Analysis Using Density-Based Spatial Clustering of Applications with Noise

Edwin Hartono — 2025-12-15

This study applies a density-based clustering method to analyze user perceptions based on reviews on Google Maps. The focus of this research lies in processing dynamic, unlabeled review data to address managers' needs in understanding public sentiment. A total of 399 data sets were collected through Apify, then the data were processed through cleaning, normalization, and stemming stages. Text representation was performed by weighting word frequencies across documents, while WordCloud visualization was utilized to identify dominant words reflecting positive perceptions to help understand the context before the clustering process. The Density-Based Spatial Clustering of Applications with Noise method was applied to form review clusters. The analysis results show that this method is able to group reviews into clusters based on content similarity, although some data were identified as noise. These findings provide useful insights in understanding public perception, thus aiding in strategic decision-making. With the right parameter selection, this method can be an effective approach for further public review sentiment analysis.

High-Performance Numerical Computation of Multidimensional Integral using Random Sampling

Andreas Widjaja — 2025-12-15

This study examines the use of high-performance computing to carry out multidimensional integral calculation based on stochastic techniques, particularly in the context of Monte Carlo integration. Considering that traditional methods are facing extreme difficulty especially in high-dimension when encountered with "dimensionality curse", random sampling technique to estimate integral values is used. This technique is superior in many aspects, for example in terms of scalability and flexibility, even in complex and irregular domains. In particular, the work concentrates on the case of calculating the volume of a multi-dimensional sphere using random sampling or Monte Carlo techniques. It also introduces a framework that employs the Graphics Processing Unit (GPU) to carry out these computations more effectively. Using dimensionalities from 2 to 24, the work compares both accuracy and computation time of the method. The results show that the random sampling method attains high accuracy in the computation of π which is used as a benchmark. The computational model is implemented in CUDA C/C++, and it takes advantage of GPU parallelism to process large sample sizes as well as execute calculations efficiently. Here it is shown in general that Monte Carlo integration is a viable approach to high-dimensional problems when combined with very rapid GPU parallelism.