Jurnal Riset Informatika

ANALYSIS OF CONTENT MANAGEMENT SYSTEM DEVELOPMENT FOR TAMAN MINI ONLINE TICKETING LANDING PAGE

Sahid Triambudhi — 2025-12-15

Taman Mini Indonesia Indah (TMII), an iconic Indonesian cultural theme park focusing on education and recreation, exhibits high promotional and event dynamics post-revitalization, necessitating its online ticket sales landing page (tiket.tamanmini.co m) to present updated information rapidly. Despite having an efficient booking system for transaction management, TMII's main landing page faces a serious operational constraint: every addition of a new ticket (ticket ID), modification, or creation of a new menu section must be executed via manual source code modification (hard code). This practice causes significant inefficiency, delays in publishing promotional tickets (e.g., school holiday bundles), and high risks of errors, directly impacting business revenue potential. This research aims to conduct a comprehensive needs analysis for designing a dedicated Content Management System (CMS) module for TMII's ticket sales landing page, thereby eliminating the reliance on hard coding. The methodology employed is qualitative descriptive, using observation and interviews with the website and operational management teams for data collection. The primary result of this analysis is a detailed specification of the functional and non-functional requirements for the CMS module, including independent CRUD (Create, Read, Update, Delete) capabilities for tickets and banners. The CMS design is expected to significantly enhance the operational efficiency of the management team, ensure content accuracy, and accelerate business response to market opportunities, ultimately making content management for online tickets independent and efficient.

TRANSFER LEARNING WITH EFFFICIENTNET-B0 FOR CAT BREED CLASSIFICATION: A COMPARATIVE EVALUATION OF OPTIMIZERS

Aini Azzah — 2025-12-15

Cats are widely kept as companion animals and exhibit substantial breed level variation in appearance and behavior that influences their care. This study develops a lightweight, image based classifier for identifying twelve common cat breeds using transfer learning on the EfficientNet-B0 backbone. Experiments contrasted four optimization algorithms (SGD, AdaGrad, RMSProp, and Adam) to identify the training strategy that balances convergence speed and generalization. Model effectiveness was measured with confusion matrix analysis and common classification indicators (accuracy, precision, recall, and F1-score). The best performing setup, EfficientNet-B0 fine tuned with the Adam optimizer attained 92% training accuracy, 89% validation accuracy, and 88% on the held out test partition. Subsequently, we integrated the trained model into a Flask web application, backed by an SQLite database, and conducted black-box testing to assess its functional reliability. All system functions met specifications and runtime predictions corresponded closely to ground truth labels. This platform provides a rapid and accurate tool for cat owners and enthusiasts to identify breeds in real-world scenarios, highlighting the usefulness of transfer learning in a streamlined web based implementation.

INFLUENCE OF LEAF IMAGING DISTANCE ON WATER GUAVA CLASSIFICATION USING NEURAL NETWORK WITH GRAY LEVEL CO-OCCURRENCE MATRIX FEATURES

Muhammad Haviz Irfani — 2025-12-15

The development of Computer Vision technology has made a significant contribution to the agricultural sector, particularly in the identification of plants based on visual characteristics. Water guava (Syzygium aqueum) is one of the fruit commodities widely cultivated in Indonesia; however, its seedling varieties are often difficult to distinguish visually. Conventional methods relying on human observation tend to have low accuracy, highlighting the need for an accurate and efficient identification system from the early stages. This study aims to analyze the effect of varying imaging distances on the extraction results of leaf vein texture features using the Gray Level Co-occurrence Matrix (GLCM) method and to evaluate how this parameter influences the classification performance of water guava seedlings using the Backpropagation Artificial Neural Network (ANN). Unlike previous GLCM–ANN plant classification studies that primarily focused on lighting or species variation, this work systematically investigates imaging distance as a key factor in optimizing texture feature stability and improving model accuracy. Experiments were conducted using five imaging distances—7 cm, 9 cm, 11 cm, 13 cm, and 15 cm—with 2,500 images used for training data and 500 images for testing data. The results show that an imaging distance of 13 cm yielded the best performance, achieving 80% accuracy, where 80 out of 100 test images were correctly classified, supported by balanced precision, recall, and F1-score values indicating stable and reliable classification performance.

DEVELOPMENT OF AN INFORMATION SYSTEM FOR ATTENDANCE AND STUDENT PROGRESS AT PAUD TUNAS MUDA

Muhammad Ghazali — 2025-12-15

The Tunas Muda Early Childhood Education Student Attendance and Progress Recording Information System application is a digital platform designed to help teachers record student attendance and progress in a modern and efficient manner. Currently, the recording process is still done manually, causing various obstacles such as late reporting, data inaccuracy, and difficulties in comprehensively monitoring student progress. This research uses the Research and Development (R&D) method. The purpose of this research is to develop a system that can facilitate teachers in taking attendance and recording student progress and enable school principals to monitor attendance and progress data through graphical displays and statistical analysis. Data collection was conducted through direct interviews with teachers as the main users. The system was developed using Flutter SDK for the interface and Firebase Firestore as the database. The results of the study show that the application is capable of recording student attendance and progress in real time, generating reports in PDF format, and displaying attendance and progress analysis in an informative graphical form.

DEVELOPMENT OF A MOBILE-BASED TRANSPORTATION AND HOTEL TICKET BOOKING INFORMATION SYSTEM AT TIKET EXTRA

Shafira Nur Alfiah — 2025-12-15

The development of information and communication technology has had a major impact on various aspects of human life, including transportation and tourism. This study aims to develop a mobile-based transportation and hotel ticket booking information system integrated through the Tiket Extra application. This application is designed to make it easier for users to search for travel schedules, compare prices, and make transactions efficiently without time and place restrictions. The research method used is Research and Development (R&D) with a Waterfall model, which includes needs analysis, system design, implementation, and testing to produce an optimal and functional system. The system is integrated with Midtrans as a payment gateway to support secure, fast, and accurate digital payment processes. Testing using the Blackbox Testing and Usability Testing methods showed that all functions worked well and were responsive. Overall, the Tiket Extra application has proven to be effective in significantly improving user convenience, agent operational efficiency, and system service quality.

CLASSIFICATION OF COFFEE LEAF SPOT DISEASES USING THE RESIDUAL NEURAL NETWORKS

Stanislaus Jiwandana Pinasthika — 2025-12-15

Coffee is one of the competitive commodities that requires detailed quality control. The common diseases that attack coffee plants are miner, rust, and phoma. Despite their visual similarity, the diseases differ in symptoms and treatments, requiring precise identification aided by computer vision. Miner and phoma have similar image features that are challenging in this study. Avoiding treatment error, several deep learning approach is needed to help classify the diseases. One of the robust methods is the Residual Network. Considering the number of datasets and alignment with the state-of-the-art, this study picked ResNet50 and ResNet101 to be observed. This study employed ResNet50 and ResNet101 in two scenarios. The first scenario was training the models on datasets without preprocessing, while the second scenario trained models on processed datasets. The preprocessing involved converting the color model to HSV and taking the range of leaf spot color from light red to dark brown for color segmentation. This study successfully achieved accuracy, precision, and F1-score at 89,16%, 89,42%, and 89,15% respectively, for the ResNet50 model trained on preprocessed data, slightly higher than the metrics of ResNet101. The ResNet101 achieved 87.95% of accuracy, 88.05% of precision, and 87.98% of F1-Score. These results indicate that ResNet50 is more robust for classifying the leaf spot, and the color segmentation helped the model to optimize the performance

APPLICATION OF TRANSFER LEARNING ON EFFICIENTNET-B0 ARCHITECTURE FOR AUTOMATIC ROOF TILE DAMAG CLASSIFICATION

Rayhan Prasetya Ady — 2025-12-15

Subjectivity in manual quality control for traditional roof tiles poses a significant challenge, as the current process relies on manual, visual inspection and subjective judgment. This research proposes an automatic system to classify tile quality from images using a Convolutional Neural Network (CNN), specifically the EfficientNet-B0 model enhanced with transfer learning. The study utilized a primary dataset comprising 616 local roof tile images collected directly from producers in Berjo Kidul, Godean, Yogyakarta. These images were manually labeled based on producer criteria into three distinct classes: 'Finished' (203 images), 'Underbaked' (213 images), and 'Broken/Cracked' (200 images). The methodology involved resizing all images to 224x224 pixels and applying data augmentation, including random rotation, horizontal flipping, and color jitter, to mitigate overfitting. The EfficientNet-B0 model, pre-trained on ImageNet, was implemented in PyTorch and trained for 10 epochs using an 80/20 train/validation split with the Adam optimizer. The model demonstrated outstanding performance, reaching 99.70% accuracy in validation. Further evaluation confirmed this robustness; the model perfectly identified the 'Underbaked' class and recorded only a single misclassification error on the test set. Qualitative analysis via a Flutter mobile application showed the system is resilient to changes in background and viewing angles, although its accuracy is compromised by poor lighting and strong shadows. This study validates the proposed system as a highly efficient and objective tool for a more reliable quality control process.

COMPARATIVE MACHINE LEARNING ALGORITHMS FOR YOUTUBE SENTIMENT ANALYSIS ON DPR DEMONSTRATION 2025 USING LEXICON

Syafri Samsudin — 2025-12-15

The high volume of public comments on YouTube regarding the DPR Demonstrasion August 2025, which reached 43,910 raw data, presents a significant challenge in conducting efficient sentiment analysis. Time and cost limitations in manual labeling for large-scale datasets are a major obstacle in the development of predictive models. This study aims to address this problem by proposing a hybrid approach that integrates Lexicon-Based auto-labeling with a comparative evaluation of five Machine Learning algorithms. The research methodology included a text preprocessing stage that generated 40,097 unique comments, feature extraction using TF-IDF, and data sharing with an 80:20 ratio. The performance of the Support Vector Machine algorithm was comprehensively compared to Random Forest, Decision Tree, K-Nearest Neighbors, and Naive Bayes. The results of the experiment showed that the SVM model recorded the most superior performance with an accuracy of 96.5% and a weighted F1-Score of 0.966. This score significantly outperformed other benchmarking algorithms, where Random Forest came in second place with 89.2% accuracy, followed by Decision Tree at 85.6%, KNN at 84.6%, and Naive Bayes at the lowest with 84.0%. These findings validate that the integration of Lexicon-Based labeling with SVM classification is a highly accurate, robust, and efficient solution for handling sentiment analysis on large-scale social media data in Indonesia.

MODELING THE IMPACT OF RECOMMENDATION ALGORITHMS ON GEN Z E-COMMERCE CONSUMPTION BEHAVIOR

Ritzqy karina — 2025-12-15

The consumptive behavior of Generation Z (Gen Z) in e-commerce platforms is strongly influenced by recommendation algorithms, which often drive impulsive purchasing decisions. This issue is further exacerbated by low levels of financial literacy and the widespread availability of Buy Now Pay Later (BNPL) services, which increase the risk of a recurring debt cycle. This study aims to model and quantitatively estimate the level of impulsive behavior using a deep learning approach. Two neural network architectures were tested and compared. The first architecture, an Artificial Neural Network (ANN), was employed as a preliminary analytical model to map the nonlinear relationships between preprocessed static variables and impulsivity levels. The second architecture, a hybrid model combining a Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM), was specifically designed to capture temporal patterns and the dynamic evolution of impulsive behavior over time. Quantitative evaluation results demonstrate that the RNN-LSTM hybrid model achieved superior performance with exceptionally high estimation accuracy, as indicated by a Mean Absolute Error (MAE) of 0.0821 and a coefficient of determination (R²) of 0.9767. In comparison, the static ANN model achieved only an MAE of 0.2078 and an R² of 0.8924. These findings explicitly confirm that impulsive behavior is a dynamic phenomenon, and thus, the hybrid RNN-LSTM architecture proves significantly more effective in analyzing sequential behavioral patterns.

SHAPE AND TEXTURE INTEGRATION FOR JAVA SEA FISH CLASSIFICATION USING K-NEAREST NEIGHBORS ALGORITHM

Pingkan Putri Nazarina — 2025-12-15

Manual identification of fish species at fish auction sites (TPI) was often time-consuming and prone to inconsistencies, which affected economic valuation and data recording accuracy. This study proposed an automated fish classification system to address these challenges using the K-Nearest Neighbors (KNN) method. The system was designed to assist the fish identification process in the Java Sea, with a case study conducted at the Karanganyar Fish Auction Site. The proposed approach employed computer vision techniques, beginning with image pre-processing steps such as segmentation and cropping to isolate fish objects. Subsequently, two complementary feature extraction methods were combined to obtain a robust representation of each fish image: Hu Moments for capturing holistic shape features that are invariant to scale and rotation, and Local Binary Pattern (LBP) for extracting detailed surface texture information. This hybrid feature representation provided a comprehensive descriptor for every fish instance. The dataset consisted of 1,000 images categorized into 10 main fish species (e.g., tongkol, bawal, and others). Model training and hyperparameter optimization were performed using a k-fold cross-validation scheme, followed by an 80:20 train-test evaluation. The experimental results demonstrated that the KNN model with the optimal k value achieved an overall classification accuracy of 98.50% on the unseen test set. These findings indicated that the integration of Hu Moments and LBP features was highly effective in distinguishing fish species and showed strong potential for practical implementation as a fast, objective, and reliable identification tool at fish auction sites such as Karanganyar Fish Auction Site

FACIAL RECOGNITION PERFORMANCE EVALUATION WITH YOLOV8, ARCFACE, AND SVM IN A CONTACTLESS EMPLOYEE ATTENDANCE SYSTEM

Glanes Cindy Terampe — 2025-12-15

Manual attendance systems, which continue to be implemented in many institutions, are vulnerable to manipulation and require significant time. This research proposes an automated facial recognition attendance system optimized to address the unique challenges posed by CCTV cameras installed at a height of 3 meters. The system integrates three main components: YOLOv8m for face detection, ArcFace for 512-dimensional feature extraction, and a Support Vector Machine (SVM) with a Polynomial kernel for identity classification. The dataset (5 classes) was augmented using 20 augmentations per image and was split into a 70% training and 30% testing ratio. An image preprocessing pipeline, including CLAHE, denoising, and sharpening, was applied to enhance the input image quality. Experimental results demonstrate high classification performance, achieving 93.7% accuracy, 0.938 precision, 0.937 recall, and an F1-Score of 0.935. Confusion matrix and PCA analysis identified that the primary misclassification occurred between the E005_employee5 and E002_employee2 classes, correlating with feature overlap. Computationally, the system achieved a throughput of 7.2 FPS on the testing hardware. The system is proven to be accurate and functional for the attendance task, although its real-time performance (FPS) is highly dependent on hardware acceleration.

DESIGN OF A DIGITAL CORRESPONDENCE AND DISPOSITION SYSTEM WITH INTEGRATED DIGITAL SIGNATURE

Kadek Rolavito Andrianto Putra — 2025-12-15

The administrative workflow at the Army Communication and Electronics Center (PUSKOMLEKAD) faces significant challenges due to its reliance on manual, paper-based correspondence processes. This manual system causes operational inefficiencies, difficulties in real-time disposition tracking, and critical workflow bottlenecks, particularly the dependency on the physical presence of leadership for signatures. Data for this study were collected through direct observation of the manual administrative workflow and interviews with personnel regarding user requirements. The research method used is Research and Development (R&D), applying the Rapid Application Development (RAD) model for the system's lifecycle using the PHP Laravel framework and MySQL database. The research resulted in a functional prototype that features an integrated digital archive, a multi-level disposition system for real-time tracking, and a secure PIN-based digital signature. In conclusion, the integration of digital signatures effectively solves the primary bottleneck by eliminating the need for physical presence, thus significantly enhancing operational efficiency, transparency, and accountability at PUSKOMLEKAD.

ANALYSIS OF CLASSIFICATION ALGORITHM IN UNBALANCED DIABETES DATASET

Ahmad Rifa'i — 2025-12-15

Diabetes mellitus is a metabolic disease that is spreading rapidly and has the potential to be life-threatening worldwide. This condition occurs when the body experiences a decline in its ability to process glucose, triggering metabolic disorders. The use of machine learning algorithms is one effective approach to predicting or detecting diabetes based on the severity of a patient's symptoms. This study uses the Diabetes dataset from Kaggle and compares the performance of several classification algorithms in unbalanced data conditions and after data balancing using the SMOTE, Random Under Sampling, Random Over Sampling, and Near Miss resampling techniques. The results show that model performance is greatly influenced by data balance conditions and the resampling method used. In the original unbalanced data condition, Artificial Neural Network (ANN) provided the best results with the highest accuracy of 96.98%, indicating that ANN is the most adaptive to class imbalance. After resampling, the performance pattern changed: with SMOTE, Random Under Sampling, and Random Over Sampling, the Random Forest algorithm consistently produced the highest accuracy of 96.52%, 89.84%, and 96.26%, respectively, demonstrating its superiority in utilizing balanced data. Meanwhile, in the Near Miss method, the best performance was achieved by Logistic Regression with an accuracy of 94.41%, indicating that minority sample selection based on proximity is more suitable for linear models. Therefore, selecting the right combination of resampling methods and machine learning algorithms is an important factor in obtaining optimal diabetes predictions.

ENHANCING SLEEP QUALITY PREDICTION THROUGH SMOTE-BASED DATA BALANCING AND HYBRID MACHINE LEARNING MODELS

Ami Rahmawati — 2025-12-15

Sleep is a vital aspect in maintaining a person's physical and psychological balance. Poor sleep quality can reduce physical and cognitive performance, increasing the risk of various health problems. This study aims to develop a predictive model for sleep quality based on factors such as lifestyle, stress, daily activities, and caffeine consumption, using XGBoost combined with Recursive Feature Elimination (RFE). XGBoost was chosen for its ability to handle imbalanced datasets and heterogeneous features, while RFE helps simplify the model without losing important information. In the data pre-processing stage, a class imbalance was found, so the Synthetic Minority Over-sampling Technique (SMOTE) process was carried out to balance the proportion of the minority class. The dataset in this study was divided into two parts, namely 80% as training data and 20% as testing data, and validated using cross-validation to ensure generalization. The results show very high model performance with an accuracy of 99.79% on training data, 99.63% on cross-validation, and 99.10% on testing data. This model was then developed into a web application for practical use in analyzing sleep quality prediction. This study emphasizes the methodological contribution of a SMOTE-based hybrid machine learning model and its ready-to-use application implementation, while also opening opportunities for further testing on more diverse datasets and evaluating biases caused by synthetic data.

SENTIMENT ANALYSIS OF MENTAL HEALTH REVIEWS USING MACHINE LEARNING ALGORITHMS

Risa Wati — 2025-12-15

Mental health is a significant issue in the modern era due to lifestyle changes, social pressures, and technological advancements that introduce new challenges. These problems affect various aspects of life, including education, employment, social relationships, and overall quality of life. Technological development enables the use of machine learning to automatically classify large amounts of data. This study aims to analyze and compare the performance of Support Vector Machines (SVM), K-Nearest Neighbor (K-NN), Naïve Bayes (NB), Logistic Regression (LR), Decision Tree (DT), and Random Forest (RF) in sentiment classification on mental health issues, while simultaneously contributing to scientific development and supporting the understanding of public psychological conditions. The dataset used in this research was obtained from Kaggle and consists of 20,364 mental health–related reviews in .CSV format, processed using Google Colab with the Python programming language. The data were categorized into two groups—depression and suicidewatch—and then underwent preprocessing, data splitting into training and testing sets with an 80:20 ratio, and TF-IDF weighting. The results indicate that the SVM algorithm outperforms the other methods. Using an RBF kernel and a C parameter of 15, SVM achieved an accuracy of 72.09%, a precision of 72.11%, a recall of 72.09%, and an F1-score of 72.09%. This study not only provides scientific contributions but also supports efforts to better understand the psychological conditions experienced by society.

IMPLEMENTATION OF A GAME RECOMMENDATION SYSTEM USING THE K-MEANS CLUSTERING AND CONTENT-BASED FILTERING METHODS

Rianggi Silvi Anti — 2025-12-15

This study focuses on developing a web-based game recommendation system using a hybrid approach, combining K-Means Clustering and Content-Based Filtering to improve the accuracy and relevance of recommendations. The dataset was taken from the RAWG API, consisting of 1,000 games with key attributes such as name, Genre, platform, rating, and age category (ESRB). The research stages included Data Preparation, exploratory analysis, attribute transformation, application of K-Means for game segmentation, and similarity calculation using Cosine Similarity. The hybrid approach was carried out by filtering recommendations based on the same cluster. The results show that the integration of the two methods produces more relevant recommendations, with UMAP and t-SNE visualizations showing clear cluster separation. The system was implemented using Django and deployed on Google Cloud Platform, resulting in an efficient, adaptive, and real-time recommendation application.

MOBILE-BASED MANAGEMENT INFORMATION SYSTEM FOR ARAFURU HOUSE MANAGEMENT

Dewi Kartika — 2025-12-15

Operational management at Wisma Arafuru currently relies heavily on conventional methods utilizing manual ledger recording. This dependence on manual processes often leads to significant administrative challenges, such as inaccuracies in guest reservation data, slow administrative handling, and difficulties for managers in monitoring room availability and facility conditions in real-time. Therefore, this study aims to design and build a mobile-based Management Information System as a digital transformation solution to enhance the operational efficiency and effectiveness of Wisma Arafuru. The methodology applied in this research is Research and Development (R&D), utilizing the systematic Waterfall software development model. The development stages include requirement analysis conducted through observation and interviews, system and interface design, application implementation using the Flutter framework supported by a Firebase backend, and comprehensive system testing. The result of this research is a mobile application that integrates key features such as room inventory management, a digital booking system, facility monitoring, and simple financial reporting. Based on implementation test results, the system has proven capable of providing convenience for managers in monitoring all operational activities, minimizing data recording errors (human error), and accelerating service processes for guests compared to the previous manual system.

BITCOIN PRICE VOLATILITY ANALYSIS: A DEEP LEARNING APPROACH TO X (FORMERLY TWITTER) SENTIMENT

Puji Astuti — 2025-12-15

This study investigates the relationship between social media sentiment and Bitcoin price volatility using advanced natural language processing techniques. We collected X data from April 10-29, 2025, analyzing cryptocurrency-related tweets alongside Bitcoin price movements obtained through the CoinGecko API. Five sentiment analysis methodologies were comparatively evaluated: VADER, TextBlob, BERTweet, RoBERTa Base, and RoBERTa Large. Bitcoin price volatility was measured using log returns to capture market fluctuations accurately. Correlation analysis revealed significant differences in methodological effectiveness. Traditional lexicon-based approaches (VADER and TextBlob) demonstrated weak correlations with volatility (r = -0.2232 and r = -0.0710 respectively). Transformer-based models showed superior performance, with RoBERTa Large achieving the strongest correlation (r = 0.4569, p = 0.0428), representing the only statistically significant relationship. The positive correlation indicates that increased social media sentiment corresponds to higher Bitcoin price volatility rather than directional price movements. These findings demonstrate that sophisticated deep learning models can effectively capture sentiment-driven market dynamics, providing valuable insights for cryptocurrency investors, trading platforms, and market analysts seeking to understand social media influence on digital asset markets.

DEVELOPMENT OF HYPEBID MARKETPLACE INFORMATION SYSTEM WITH REAL-TIME ONLINE AUCTION FEATURE

Jovan Avriananta — 2025-12-15

Conventional auctions still face a number of challenges, such as limited access, unclear processes, and low time efficiency. Auction participants are generally required to be physically present, which limits the number of participants and reduces transparency and transaction speed. Based on this background, this study aims to develop HypeBid, a web- and mobile-based marketplace information system that supports online and real-time auction processes. This system is built using a client-server architecture with React Native, Express.js, PostgreSQL, and Supabase technologies. Development was carried out through stages of needs analysis, system design, and implementation of key features such as user registration, product verification, live bidding, integrated payment systems, and transaction reports. System testing was conducted using the black box testing method involving two groups of users, namely buyers and auction officers. The test results showed that all features functioned as expected, without any functional errors. Thus, HypeBid is considered to be a clearer, more flexible, and efficient alternative solution compared to conventional auction methods.