JSAI (Journal Scientific and Applied Informatics)

Sentiment Analysis of the MPStore Application Using Logistic Regression and LDA Algorithms

Tia Arlin Dita, Ali Ibrahim, Rizka Rahmadhani, Mira Afrina — Tue, 30 Dec 2025 00:00:00 +0700

The rapid growth of the digital economy encourages user satisfaction as the key to successful application innovation. Within technopreneurship, understanding user sentiment is essential for sustainable product development. This study aims to analyze sentiment and identify the deter-minants of user satisfaction regarding the MPStore application based on reviews from the Google Play Store. Review data were collected via scraping and analyzed using Logistic Regression (LR) for sentiment classification (positive, negative, neutral) also Latent Dirichlet Al-location (LDA) for satisfaction topic extraction. The result shows that the LR model achieved an accuracy of 88.5%. The LDA analysis also successfully revealed eight main topics, includ-ing ease of use, transaction speed, and technical obstacles (errors, login, balance issues). Over-all, a majority of users hold a positive perception of MPStore's efficiency and ease of transac-tions. This study concludes that the combination of sentiment analysis and topic modeling is effective for explaining the level of user satisfaction and providing a strategic foundation for digital application developers.

Retrieval-Augmented Generation Method in the Development of Large Language Model Chatbots for the Anambas Civil Registry Public Information Service

Muhammad Habsyi Mubarak, Joko Sutopo — Tue, 30 Dec 2025 00:00:00 +0700

Population administration services in the Anambas Islands Regency face significant challenges related to limited information access caused by geographical conditions. Service information on the official Disdukcapil website, which is passive, often makes it difficult for the public to obtain fast and relevant answers. This condition leads to service queue buildups and potentially decreases public satisfaction. As a solution, an intelligent chatbot application based on a Large Language Model (LLM) with a Retrieval-Augmented Generation (RAG) approach was developed. This method effectively combines precise information retrieval capabilities from a document database with the LLM's natural language understanding ability to produce contextual answers. The system's development process was carried out using the LangChain framework, Chroma vector store, and was integrated into a web interface as the frontend. Official Disdukcapil service information was processed through chunking, embedding, and RAG pipeline creation stages. The results showed that the chatbot was able to respond to inquiries about population services accurately and efficiently. Based on evaluations using the BERTScore metric, the system obtained average scores of 98,5% for Precision, 99,1% for Recall, and 98,8% for F1-Score. This system can be accessed from anywhere, greatly assisting people in remote areas, and serves as a potential initial prototype to support the digitization of public services in archipelago regions.

Information Technology, Social Media, and Digital Transformation on the Performance of Laundry MSMEs in Batam City

Surya Tjahyadi, Cindy Claudia Erica, Hendi Sama — Tue, 30 Dec 2025 00:00:00 +0700

This study examines the effects of information technology, social media, and digital transformation on the business performance of laundry MSMEs in Batam City using a quantitative associative approach. Data were collected from 94 owners or managers of laundry MSMEs through questionnaires and analyzed using IBM SPSS Statistics 25. Classical assumption tests indicated that the data were normally distributed and free from multicollinearity and heteroscedasticity. Simultaneous testing showed that information technology, social media, and digital transformation jointly have a significant effect on business performance (F = 4.588; p = 0.005). However, partial test results revealed that only social media has a positive and significant effect on business performance, while information technology and digital transformation do not show significant effects. These findings suggest that in small-scale service MSMEs, market-oriented digital tools are more effective in improving business performance than complex internal technology adoption.

Comparative Analysis of UI/UX of Online Transportation Booking Applications Using the System Usability Scale (SUS) and Heuristic Evaluation

Mon, 12 Jan 2026 00:00:00 +0700

This study aims to analyze and compare the User Interface (UI) and User Experience (UX) of three online transportation applications in Palembang City, namely Gojek, Grab, and Maxim. The rapid growth of Indonesia’s online transportation industry, which reached 88.3 million users in 2024, highlights the importance of evaluating usability quality. Palembang has unique geographical characteristics, as the city is divided by the Musi River, creating specific challenges for app-based transportation services. The novelty of this study lies in the integration of the System Usability Scale (SUS) with Nielsen’s heuristic evaluation within an urban geographical context fragmented by natural barriers. A descriptive comparative approach was employed, involving 120 active users and three UI/UX expert evaluators. The results show that Gojek achieved the highest SUS score of 78.5 (Good), followed by Grab with 75.2 (Good) and Maxim with 68.7 (OK). One-Way ANOVA indicates significant differences among the applications (F = 12.847; p < 0.001). Theoretically, these findings demonstrate that an integrated usability evaluation approach provides a more comprehensive understanding of UI/UX quality in geographically specific contexts.

Development and Evaluation of the Usability of the Hadroh Instrument Marketing Website Using the System Usability Scale (SUS)

Danu Dwi Setiawan, Akhmad Fadjeri — Mon, 12 Jan 2026 00:00:00 +0700

The rapid growth of digital technology has encouraged micro, small, and medium enterprises (MSMEs) to adopt digital marketing strategies in order to remain competitive. This study aims to develop a marketing website for the Syekh Hadroh MSME and to evaluate the quality of the developed system in terms of functionality and usability. The research employed a Research and Development (R&D) approach with the Waterfall model as the system development method. Functional testing was conducted using Black Box Testing with the Equivalence Partitioning technique, while usability evaluation was performed using the System Usability Scale (SUS) involving 10 respondents. The results of Black Box testing indicate that all test scenarios were successfully executed with a success rate of 100%, demonstrating that the system functions according to the specified requirements. Furthermore, the SUS evaluation produced an average score of 73, which falls into the “good” usability category, indicating that the system is easy to use and well accepted by users. These results suggest that the developed marketing website is functionally reliable and usable, and therefore suitable for implementation as a digital marketing medium to support market expansion and improve service quality for MSMEs.

Analysis of the Effect of Conventional QRIS and ESB QRIS Use on Digital Payment Efficiency at BJ Restaurants in the Digital Age

Wishnu Aribowo Probonegoro, Lili Indah Sari, Ardiana — Mon, 12 Jan 2026 00:00:00 +0700

The development of digitalization has encouraged the adoption of cashless payment systems through the Quick Response Code Indonesian Standard (QRIS) in the culinary sector, including Rumah Makan BJ. This study aims to analyze the effect of Conventional QRIS and ESB QRIS on digital payment efficiency. This research employs a quantitative approach with a sample of 122 respondents. Data were collected through questionnaires and analyzed using SPSS version 17. The partial test results indicate that Conventional QRIS has a positive and significant effect on digital payment efficiency, with a t-value of 2.802 and a significance level of 0.006 (<0.05). ESB QRIS shows a stronger influence, with a t-value of 6.029 and a significance level of 0.000 (<0.05). Simultaneously, the ANOVA test shows an F-value of 57.155 with a significance of 0.000, indicating that both variables jointly have a significant effect on digital payment efficiency. The coefficient of determination (R²) of 0.405 indicates that 40.5% of the variation in digital payment efficiency is explained by QRIS usage. These findings are consistent with previous studies on integrated digital payment systems.

Modeling and Simulation of Car Robot Path Planning Using the Ant System Method in a Structured Environment

Rama Saktriawindarta, Suhendri Suhendri, Nurita Evitarina — Tue, 20 Jan 2026 00:00:00 +0700

Path planning is a fundamental problem in mobile robot navigation that requires efficient route optimization while avoiding obstacles. This study implements the Ant System method to solve the mobile robot path planning problem using a modeling and simulation approach in a structured environment. The path planning process utilizes pheromone mechanisms and heuristic information to determine an optimal path from the initial state to the goal state. A total of ten simulation experiments were conducted with variations in the number of intermediate coordinates, iterations, and ants. The results show that the proposed method successfully generated collision-free paths in all experiments, with path lengths ranging from 31.4915 to 32.6788 units. The analysis indicates that the balance between the number of intermediate coordinates, iterations, and ants significantly affects path quality, where well-balanced parameters produce smoother and more stable trajectories. Overall, the Ant System method achieved a 100% success rate, demonstrating its effectiveness and reliability for mobile robot path planning in structured environments.

Decision Tree Modeling Using the ID3 Algorithm in a Data Mining Approach

Suhendri Hendri, Rama Saktriawindarta, Nurita Evitarina — Tue, 20 Jan 2026 00:00:00 +0700

The rapid development of information technology has encouraged the use of data mining as a foundation for data-driven decision-making across various sectors, including the karaoke entertainment industry. This study aims to evaluate the performance of the ID3 algorithm in supporting decision support systems through the construction of a decision tree–based classification model. The research method employs the Knowledge Discovery in Databases (KDD) approach, which involves data selection, data transformation, modeling using the ID3 algorithm, and evaluation of decision outcomes. The performance of the method was evaluated based on five key aspects: decision-making capability, classification processing speed, classification result stability, model interpretability, and suitability to user needs. The results indicate that the ID3 algorithm achieved an average success rate of 92%, with the highest performance observed in processing speed and classification stability. These findings demonstrate that the ID3 algorithm is effective, efficient, and highly interpretable, making it suitable for implementation as a classification method in data mining–based decision support systems.

Predicting Flood Potential Using Machine Learning with the XGBoost and Logistic Regression Approaches

Nurita Evitarina, Fitriyanti Fitriyanti, Tri Dewi Yuni Utami — Tue, 20 Jan 2026 00:00:00 +0700

Flooding is one of the most frequent natural disasters in Indonesia, causing significant material losses and casualties. This study aims to develop a flood potential prediction model based on weather data using machine learning approaches, namely XGBoost and Logistic Regression. The dataset consists of 1,513,505 weather records with 1,165 flood events (0.077%). The features include temperature, humidity, wind speed and direction, weather codes, and temporal features generated using a sliding window approach for H-1, H-2, and H-3. Data imbalance was addressed using a combination of stratified undersampling and SMOTE, changing the class ratio from 1:1,298 to 1:3.3. Experimental results show that XGBoost outperforms Logistic Regression, achieving an accuracy of 98.40%, precision of 97.93%, recall of 95.07%, and an ROC-AUC of 99.38%, while Logistic Regression achieved an accuracy of 62.77%. Feature importance analysis indicates that weather codes at H-3 and H-1 are the most influential predictors. With a low false negative rate of 4.9%, the proposed XGBoost model is considered reliable for implementation as a flood early warning system.

Causal Model of the Influence of Digital Facilities and Content on Students' Information Technology Competence with Learning Interest Mediation

Ezar Aziz, Yulianti, Wishnu Aribowo Probonegoro — Fri, 23 Jan 2026 00:00:00 +0700

This study aims to analyze the effect of facility availability and digital content on students’ information technology competencies with learning interest as a mediating variable. A quantitative approach with a survey method was employed involving 178 students from Information Technology and Visual Communication Design programs. Data were analyzed using multiple regression, simple regression, and path analysis with SPSS 25. The results indicate that facility availability and digital content have a positive and significant effect on students’ learning interest, accounting for 68.2% of the variance. Learning interest also has a significant positive effect on students’ information technology competencies, contributing 61.5% of the variance. Furthermore, digital content shows a significant indirect effect on students’ competencies through learning interest. These findings highlight the strategic role of learning interest as a mediating variable in the causal model of digital learning and imply that improving students’ competencies requires not only adequate facilities and high-quality digital content but also instructional strategies that foster students’ learning interest.

Application of Naive Bayes Algorithm with TF-IDF Weighting and Lexicon Approach for Sentiment Analysis of Student Opinions on Campus Facilities

Honainah — Fri, 23 Jan 2026 00:00:00 +0700

This study aims to analyze student sentiment towards campus facilities at Nurul Jadid University using the NaiveBayes method. Data was collected through questionnaires and processed using Visual Studio Code software. The process stages included manual sentiment classification, data preprocessing, lexicon classification, TF-IDF weighting, and classification using the Naive Bayes method. The results of each step are presented in tables and graphs. The system was also implemented by creating a web application that allows users to enter new text/opinions and obtain sentiment classification results. The results of testing manual sentiment classification and lexicon with the Naive Bayes algorithm showed different levels of accuracy, with manual sentiment classification having an accuracy value of 75% and lexicon sentiment classification having an accuracy value of 85%. In conclusion, the lexicon approach with the Naive Bayes algorithm is superior to the manual approach because it is more objective, consistent, efficient, and easy to develop and is suitable for analyzing student opinions on campus facilities, thus providing material for consideration in campus policy.

Evaluation of the Development of an Automated Workflow-Based New Student Admission Information System Using a Research and Development Approach

Fri, 23 Jan 2026 00:00:00 +0700

The New Student Admissions (PMB) process is a strategic endeavor that influences the quality of university entrants. Nonetheless, in several colleges, particularly in remote regions, the execution of PMB remains manual, which may lead to issues regarding efficiency, accuracy, and openness in data management. This study seeks to create and assess an automated workflow-oriented PMB information system to enhance PMB management at the Muhammadiyah Selayar Institute of Technology, Science, and Business. The system development employs the Waterfall methodology, encompassing the phases of requirements analysis, design, implementation, and system testing. Evaluation is conducted by assessing the validity, practicality, and efficacy of the system with the participation of internal users as evaluators. The evaluation results demonstrated a validity rate of 100%, practicality of 95.83%, and effectiveness of 89.56%, demonstrating the system's feasibility for supporting the PMB process. This system can systematically combine the registration process, document verification, selection, and results announcement. This research contributes an automated workflow implementation model within the PMB information system, enhancing process management efficiency and facilitating the oversight of PMB activities. While the test remains confined to the institution's internal setting, the findings of this study are anticipated to serve as a benchmark for the establishment of a comparable PMB system in universities with analogous attributes.

Implementation of Rapid Application Development and Evaluation of User Experience Questionnaire in Website-Based PKL Registration Information System at UMNU Kebumen

Rifki Ardiansah, Ghufron Zaida Muflih — Fri, 23 Jan 2026 00:00:00 +0700

Field Work Practice (PKL) as a mandatory curriculum component at Ma'arif Nahdlatul Ulama University (UMNU) Kebumen still faces administrative challenges such as data duplication, non-real-time quota validation, and non-integrated payment verification. Unlike similar systems that focus only on basic registration, this study develops a web-based PKL registration information system by integrating real-time quota validation features and a payment verification mechanism by the treasurer. The system was developed using the Rapid Application Development (RAD) method to accelerate the development process. The system was built with the Laravel framework and MySQL. Functional testing results show that all features run with 100% accuracy. Evaluation using the User Experience Questionnaire (UEQ) on 42 respondents yielded positive scores on six aspects: attractiveness (1.623), clarity (1.524), efficiency (1.655), accuracy (1.589), stimulation (1.601), and novelty (1.244). The system has been proven to optimize the PKL administrative process and can serve as a model for developing integrated systems in higher education institutions.

Evaluation of Web-Based Crystal Ice Distribution Information System Using System Usability Scale (SUS) and User Acceptance Testing (UAT)

Rena Judha Wijayanti, Sri Dianing Asri — Fri, 23 Jan 2026 00:00:00 +0700

This study aims to evaluate the usability and user acceptance of a web-based crystal ice distribution information system developed to support the operations of PT Eshokita Tubindo Mandiri. The system was developed using the Waterfall method with stages of requirements analysis, design, implementation, and testing. The system evaluation was conducted using the System Usability Scale (SUS) and User Acceptance Testing (UAT) on 13 respondents who were directly involved in the distribution process. The test results showed a SUS score of 82.11, which is classified as very good, and all system functions were accepted by users based on the UAT results. However, the scope of testing was still limited to one company and a relatively small number of respondents. This study contributes to the evaluation of cold chain distribution information systems based on usability and user acceptance.

Evaluation of Retrieval Methods in Domain-Specific Chatbots Based on Retrieval-Augmented Generation

Asmaidin Asmaidin, Cahyono Budy Santoso — Fri, 23 Jan 2026 00:00:00 +0700

This study evaluated retrieval methods in the implementation of a domain-specific chatbot based on Retrieval-Augmented Generation to improve information accuracy and relevance while reducing hallucination risks. The primary problem addressed was the incorrect selection and prioritization of contextual documents in chatbot systems built on large language models, particularly in technical domains. An experimental approach was applied by comparing three retrieval strategies: lexical retrieval based on term frequency–inverse document frequency, semantic retrieval using vector representations, and a hybrid retrieval method combining lexical and semantic signals. System performance was measured using Recall at different ranking thresholds and Mean Reciprocal Rank to assess both document discovery and ranking quality. The results demonstrated that lexical retrieval achieved the highest precision at the top-ranked position, while semantic retrieval showed reduced effectiveness due to semantic drift in technical documents. The hybrid approach improved mid-range recall performance but still exhibited ranking ambiguity for top-ranked results. These findings indicated that retrieval quality in Retrieval-Augmented Generation systems depended more on effective ranking and context prioritization than on document availability alone. The study concluded that systematic evaluation of retrieval methods was essential for developing reliable domain-specific chatbots.

Evaluation of the Success of Regional Trade Information System Implementation Based on the HOT-Fit Framework and the Technology Acceptance Model

Dzalfa Tsalsabila Rhamadiyanti, Aditya Ahmad Fauzi, Fithriawan Nugroho — Fri, 23 Jan 2026 00:00:00 +0700

The Pangkalpinang City Trade Information System (SIPGK) was developed as a digital instrument to support trade data management and data-driven public information services. This study aims to evaluate the implementation success of SIPGK using the Human–Organization–Technology Fit (HOT-Fit) model, with the Technology Acceptance Model (TAM) employed as a complementary interpretative lens. A qualitative evaluative approach was applied through observation, interviews, and system documentation. The results indicate that the technology aspect demonstrates a system availability rate of 95%, reflecting good system quality and service stability, while the organizational aspect is supported by formal policies and standard operating procedures. However, the human aspect remains a key limiting factor due to disparities in digital literacy and data input consistency, along with suboptimal cross-unit data integration. These findings reveal a gap between technological and organizational readiness and human resource capacity in achieving strategic system utilization. The novelty of this study lies in applying the HOT-Fit model to a regional trade information system context, which has been rarely examined, and in integrating TAM as an interpretative framework to explain user acceptance.

Implementation and Performance Evaluation of ESP32-Based Multi-Sensor IoT System for Real-Time Environmental Monitoring and Early Warning

Fri, 30 Jan 2026 00:00:00 +0700

Real-time environmental monitoring has become increasingly important due to growing urban and industrial activities that affect air quality, noise levels, and physical environmental stability. However, many existing monitoring systems remain relatively expensive, lack portability, and are limited to passive monitoring functions without clear performance evaluation. This study aims to implement and evaluate the performance of an Internet of Things (IoT)-based multi-sensor environmental monitoring system integrated with a mobile application and real-time early warning features. The system is developed using an ESP32 microcontroller connected to DHT22, MQ135, SW-420, and KY-037 sensors to monitor temperature, humidity, air quality, vibration, and noise levels. Sensor data are transmitted to a server via a RESTful API, stored in a MySQL database, and visualized in real time through a Flutter-based mobile application. The research adopts a Research and Development (R&D) approach, encompassing requirement analysis, system design, implementation, integration, and functional testing. The experimental results indicate that the system can transmit multi-sensor data reliably with low response time, present environmental information in real time, and consistently deliver early warning notifications when environmental parameters exceed the defined threshold values. This study contributes by providing a practical and replicable performance evaluation of an IoT-based multi-sensor system suitable for small-scale environmental monitoring.

Deep Learning-Based Multi-Class Waste Classification Using the VGG16 Model

Vina Ayumi — Fri, 30 Jan 2026 00:00:00 +0700

The manual waste sorting process has faced various challenges, such as low efficiency and a high potential for classification errors. This study aimed to implement and analyze the performance of a deep learning–based VGG16 model for multi-class waste classification using digital images. The dataset used consisted of six waste classes, namely cardboard, glass, metal, paper, plastic, and residual waste, with an imbalanced initial number of images. To address this issue, data augmentation was performed so that each class contained 500 images. The dataset was then divided into 70% training data, 15% validation data, and 15% testing data. The experiments were conducted using a transfer learning approach by varying training parameters, including the RMSProp, Adam, and Stochastic Gradient Descent (SGD) optimizers, as well as batch sizes of 16, 32, and 64. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. The results showed that the selection of training parameters significantly affected model performance. The best configuration was achieved using the VGG16 model with the Adam optimizer and a batch size of 16, which produced the highest testing accuracy of 85.87%. This study was expected to serve as a foundation for the development of automated computer vision–based waste sorting systems

Usability Analysis in the Design of a Web-Based Warehouse Asset Management Information System Using the Prototyping Method and System Usability Scale (SUS)

rubiyatno, Sri Dianing Asri — Fri, 30 Jan 2026 00:00:00 +0700

Operational asset management plays a strategic role in ensuring the smooth and continuous operation of warehouse activities within a company. PT. Indomarco Prismatama faces challenges in managing assets, particularly in the processes of asset borrowing, return, and maintenance, which are still conducted manually. This results in potential delays in data recording, reduced data accuracy, and limited information transparency. This study aims to design a web-based information system for managing asset borrowing, return, and maintenance to improve efficiency and accuracy. The system was developed using the Software Development Life Cycle (SDLC) methodology with a prototyping approach, which allows for user involvement through iterative stages. The system was built using the CodeIgniter 4 framework and a MySQL database. System testing was performed using the System Usability Scale (SUS) method, involving 20 respondents, including Asset Admins, VUM Admins, and Managers. The evaluation results show an average SUS score of 81.9, indicating a high level of usability (Excellent). However, these results are limited to the current group of respondents and may not represent a broader user base. Overall, the developed system supports more effective, efficient, and transparent operational asset management, although further testing is needed to confirm its scalability and usability across a wider audience.

Application of ResNet50 for Oil Palm Fruit Image Classification Based on Ripeness Level

Hadiguna Setiawan, Handrie Noprisson, Abraham Cornelius Dachi, Ilim Hilimudin — Fri, 30 Jan 2026 00:00:00 +0700

Manual ripeness assessment still has limitations as it is subjective and highly dependent on human expertise. Therefore, this study aims to apply a deep learning approach based on the ResNet50 architecture to classify oil palm fruit ripeness into three categories, namely unripe, ripe, and overripe. The dataset used in this study consists of 1,350 RGB images of oil palm fruits, which are divided into training, validation, and testing sets with a ratio of 70:10:20. All images are preprocessed by resizing them to 224 × 224 pixels and normalizing pixel values, while data augmentation is applied to the training set to improve model generalization. A pre-trained ResNet50 model on the ImageNet dataset is employed as a feature extractor and trained using the Adam optimizer with a learning rate of 1 × 10⁻⁴ for 50 epochs. Experimental results show that the model achieves an accuracy of 89.7% on the training data and 84.1% on the validation data. Evaluation on the testing data yields an accuracy of 84.07%, with average precision, recall, and F1-score values of 84.71%, 84.07%, and 84.32%, respectively. These results indicate that the proposed ResNet50-based model demonstrates good and stable performance in classifying oil palm fruit ripeness levels.

Contrastive Learning on IndoBERT for Sentiment Analysis of Free Nutritious Meal Policies

Dwi Dian Sari Nonibenia Hia, Sunneng Sandino Berutu, Jatmika — Fri, 30 Jan 2026 00:00:00 +0700

Transformer-based language models such as IndoBERT still face limitations in topic and sentiment analysis of short social media texts, particularly due to embedding anisotropy, semantic overlap between topics, and limited sensitivity to implicit sentiment intensity. This study aims to evaluate the effectiveness of integrating SimCSE-based contrastive learning to optimize IndoBERT vector representations for sentiment analysis of the “Free Nutritious Meals” public policy. A comparative experimental approach was employed using an equal number of topics (three topics) and evaluated through BERTopic and Aspect-Based Sentiment Analysis (ABSA). The results demonstrate that the contrastive learning–based model substantially improves cluster separability, indicated by an increase of more than 1000% in the Silhouette Score compared to the baseline model, along with a reduction in topic overlap of approximately 40–50%. In addition, topic keyword diversity increased by more than 75%, yielding more informative and interpretable topic representations. In aspect-based sentiment analysis, the contrastive model exhibited approximately a 50% improvement in sensitivity to sentiment intensity and achieved perfect classification of implicit high-confidence sentiments that were previously misclassified as neutral by the baseline model. These findings confirm that contrastive learning–based embedding optimization effectively addresses the limitations of conventional embeddings and enhances the quality of topic modeling and aspect-based sentiment analysis for Indonesian social media texts.

Plant Disease Recognition Based on Leaf Images Using Sequential-Based DenseNet Architecture

Mariana Purba — Fri, 30 Jan 2026 00:00:00 +0700

Plant diseases that affect leaves can significantly reduce crop quality and productivity, making accurate and efficient detection methods essential. This study aims to develop a plant disease recognition model based on leaf images using a sequential DenseNet121 architecture. The dataset consists of 1,530 leaf images categorized into three classes: Healthy, Powdery, and Rust, which are divided into training, validation, and testing sets with a relatively balanced distribution. The model employs DenseNet121 as a base model with pre-trained ImageNet weights, where all base layers are frozen to function as a feature extractor. The classification process utilizes GlobalAverage Pooling2D, Dense, Dropout, and Softmax layers. Experimental results show that the model achieves an accuracy of 98.28% on the training data and 96.25% on the validation data. Evaluation on the test dataset yields an accuracy of 93.33%, indicating that the proposed model demonstrates good generalization capability in classifying plant diseases based on leaf images. These results suggest that the sequential DenseNet architecture is effective for plant disease recognition and has potential for further development as a decision support system in agriculture

LSTM Algorithm Analysis for Opinion Classification on the Development of Oil Palm Plantations in Indonesia

Hadiguna Setiawan, Handrie Noprisson, Abraham Cornelius Dachi, Ilim Hilimudin — Fri, 30 Jan 2026 00:00:00 +0700

This study aims to analyze public opinion on the development of oil palm plantations in Indonesia through sentiment classification using the Long Short-Term Memory (LSTM) algorithm. The data used in this study were taken from Twitter by collecting 750 tweets consisting of three sentiment categories: positive, negative, and neutral. The pre-processing stage includes filtering, tokenization, stemming, and word-embedding to prepare the data for further analysis. The LSTM model was applied to classify the sentiment of the processed tweets, and evaluated using accuracy, precision, recall, and F1-score metrics. The evaluation results showed that the LSTM model produced an accuracy of 70.81%, with precision, recall, and F1-score varying between classes, namely 0.92, 0.71, and 0.80 for the negative class, 0.48, 0.63, and 0.55 for the neutral class, and 0.77, 0.77, and 0.77 for the positive class. This study shows that LSTM can be used to analyze public opinion on the issue of oil palm plantations, despite challenges in classifying neutral tweets.

Evaluation of the Success of the Integrated Attendance and Payroll Information System Based on the DeLone and McLean Model

Imam Tauzy, Sri Dianing Asri — Fri, 30 Jan 2026 00:00:00 +0700

Employee attendance and payroll systems play a crucial role in human resource management as they directly affect payroll accuracy, transparency, and employee satisfaction. However, many previous studies have primarily focused on system development and functional testing, with limited attention to comprehensive system success evaluation. This gap highlights the need for an evaluative approach that measures system quality, information quality, and the benefits perceived by users. This study aims to evaluate the success of an integrated web-based attendance and payroll information system using the DeLone and McLean Information System Success Model. The research employed a quantitative approach with a survey method, using a Likert-scale questionnaire distributed to system users. Data were analyzed using descriptive quantitative analysis and converted into percentage values. The results indicate that the system achieved a success rate of 92%, categorized as very good. These findings demonstrate that the system has high system quality and information quality and provides significant net benefits in improving administrative efficiency and payroll information transparency for employees.

Application of Convolutional Neural Network Optimization for Multi-Class Classification of Brain Tumors in MRI Images

Vina Ayumi — Fri, 30 Jan 2026 00:00:00 +0700

Brain tumors are among the most critical neurological diseases and require early and accurate diagnosis to support appropriate medical treatment. Magnetic Resonance Imaging (MRI) is widely used for brain tumor detection due to its high-resolution imaging capability; however, manual analysis of MRI images is time-consuming and highly dependent on the expertise of radiologists. Therefore, this study aims to apply an optimized Convolutional Neural Network (CNN) for multi-class brain tumor classification using MRI images. The dataset used in this study consists of 7,023 MRI images, categorized into four classes: glioma, meningioma, pituitary, and healthy, and divided into training, validation, and testing subsets. The research stages include image preprocessing, CNN architecture design, hyperparameter optimization, model training for 50 epochs, and performance evaluation. The training process achieved an accuracy of 87.44%, while the validation accuracy reached 85%, indicating good model generalization. Model evaluation on the test dataset using a confusion matrix, precision, recall, F1-score, and accuracy resulted in an overall accuracy of 77.8%.

Classification of Rice Leaf Diseases and Pests Using the ResNet50 Model on the AgroGuard AI Dataset

Mariana Purba — Fri, 30 Jan 2026 00:00:00 +0700

Rice leaf diseases and pests are one of the main factors causing decreased rice productivity. Manual disease identification still relies on the experience of farmers and extension workers, potentially leading to delayed diagnosis and mishandling. This study aims to develop an image-based rice leaf disease and pest classification model using the ResNet50 deep learning architecture. The dataset used comes from AgroGuard AI and consists of seven classes: blast disease, healthy leaves, insect attacks, leaf roller pests, leaf scald disease, brown spot disease, and tungro disease. The dataset is divided into training, validation, and test data with a ratio of 70%:15%:15%, where the test data is balanced with 400 images in each class. The ResNet50 model was trained from scratch without pre-training weights with a batch size of 32, a learning rate of 0.001, and 50 epochs. The evaluation results showed that the model achieved an accuracy of 77.86% on the test data, with a training accuracy of 80.52% and a validation accuracy of 89.38%. Evaluation using a confusion matrix and precision, recall, and F1-score metrics indicated that the model performed quite well and stably across all classes.

Evaluation of Makeup Artist Service Ordering System Using Rapid Application Development and PIECES Methods

Jhon Cavlin Jovin, Wawan Kurniawan — Fri, 30 Jan 2026 00:00:00 +0700

This study aims to evaluate a Make Up Artist service booking system using the Rapid Application Development (RAD) method with PIECES evaluation. The evaluation was conducted using a pre-test and post-test approach to compare system conditions before and after implementation. The PIECES method was applied to measure system quality based on six dimensions: performance, information, economy, control, efficiency, and service. The results indicate that the average system quality score increased from 54% before implementation to 84% after implementation, showing an improvement of 30%. These findings demonstrate that the implemented system significantly improves service quality, operational efficiency, and customer satisfaction. Therefore, the combination of RAD and PIECES evaluation is effective in enhancing the quality of Make Up Artist service booking systems.

Music Recommendation Model Based on Semantic Representation of Song Lyrics Using BERT

Dziaul Hululiah zia, Fersellia Fersellia — Fri, 30 Jan 2026 00:00:00 +0700

The rapid growth of digital music platforms has resulted in an information overload problem, making it difficult for users to discover songs that match their preferences. This study proposes a content-based music recommendation model through semantic analysis of song lyrics using a Natural Language Processing approach with Bidirectional Encoder Representations from Transformers. The research stages include Indonesian song lyric data collection, data cleaning, text preprocessing, contextual lyric embedding generation, and lyric similarity computation using cosine similarity. Model performance is evaluated using Mean Squared Error and accuracy. Experimental results show that the proposed model achieves an accuracy of 83.69% with a Mean Squared Error value of 1.4066, indicating that lyric representations generated by Bidirectional Encoder Representations from Transformers effectively capture semantic meaning and quantitatively improve the relevance of music recommendations. Therefore, the proposed approach enhances the accuracy and personalization of content-based music recommendation systems.

Linear Regression Analysis and Feature Contribution-Based Ensemble Learning in Electric Car Price Prediction

Nur Oktavin Idris, Fuad Pontoiyo — Fri, 30 Jan 2026 00:00:00 +0700

This study aims to analyze the performance of linear regression and ensemble learning methods in predicting electric vehicle prices based on technical specifications, as well as to examine the contribution of key features to the prediction results. The main challenge in electric vehicle price prediction lies in the high price variability driven by nonlinear relationships among technical attributes, which are difficult to capture using simple linear models. Linear regression was employed as a baseline model, while Random Forest and Gradient Boosting were used as ensemble learning approaches. The dataset was obtained from Kaggle and processed through data cleaning, categorical encoding, normalization, and an 80:20 train–test split. Model performance was evaluated using mean squared error (MSE) and the coefficient of determination (R²). The results indicate that the Gradient Boosting model achieved the best performance, with an MSE of 8.63 and an R² of 0.891, outperforming both Random Forest and linear regression models. Feature contribution analysis reveals that vehicle acceleration time is the most influential factor in determining electric vehicle prices. These findings demonstrate that ensemble learning not only improves predictive accuracy but also provides analytical insights into the key technical factors shaping electric vehicle pricing.