Jurnal Teknik Informatika (JUTIF)
Not a member yet
992 research outputs found
Sort by
Early Fusion of CNN Features for Multimodal Biometric Authentication from ECG and Fingerprint Using MLP, LSTM, GCN, and GAT
Get PDFTraditional authentication methods such as PINs and passwords remain vulnerable to theft and hacking, demanding more secure alternatives. Biometric approaches address these weaknesses, yet unimodal systems like fingerprints or facial recognition are still prone to spoofing and environmental disturbances. This study aims to enhance biometric reliability through a multimodal framework integrating electrocardiogram (ECG) signals and fingerprint images. Fingerprint features were extracted using three deep convolutional networks—VGG16, ResNet50, and DenseNet121—while ECG signals were segmented around the first R-peak to produce feature vectors of varying dimensions. Both modalities were fused at the feature level using early fusion and classified with four deep learning algorithms: Multilayer Perceptron (MLP), Long Short-Term Memory (LSTM), Graph Convolutional Network (GCN), and Graph Attention Network (GAT). Experimental results demonstrated that the combination of VGG16 + LSTM and ResNet50 + LSTM achieved the highest identification accuracy of 98.75 %, while DenseNet121 + MLP yielded comparable performance. MLP and LSTM consistently outperformed GCN and GAT, confirming the suitability of sequential and feed-forward models for fused feature embeddings. By employing R-peak-based ECG segmentation and CNN-driven fingerprint features, the proposed system significantly improves classification stability and robustness. This multimodal biometric design strengthens protection against spoofing and impersonation, providing a scalable and secure authentication solution for high-security applications such as digital payments, healthcare, and IoT devices
Integration of Squeeze-and-Excitation in Densenet-121 for Classifying Real and AI-Generated Images
Get PDFRecent advancements in generative technologies, such as Generative Adversarial Networks (GANs) and Latent Diffusion Models, have enabled the creation of AI-generated synthetic images that are increasingly indistinguishable from real ones, posing significant challenges for verifying the authenticity of visual content. This study develops a DenseNet-121 model with hyperparameter optimization and the integration of Squeeze-and-Excitation (SE) attention mechanisms at Early, Mid, and Late positions. Experiments were conducted using the CIFAKE dataset with a resolution of 32×32 pixels to compare the baseline Plain model with three SE variants. Hyperparameter optimization was applied to maximize model performance. The results demonstrate that the Plain DenseNet-121 with optimized hyperparameters achieved an accuracy of 98.52%, outperforming the standard configurations reported in previous studies. The integration of SE yielded varied outcomes, where Mid SE attained the highest accuracy of 98.56%, while Early SE (98.45%) and Late SE (98.48%) exhibited greater stability with lower standard deviations. These findings highlight that combining hyperparameter optimization with appropriate SE placement can enhance model performance for classifying real and AI-generated images. Moreover, SE placement at different positions (Early, Mid, Late) has a significant impact on feature representation and generalization in synthetic image classification, which is increasingly important given the growing difficulty of distinguishing real from AI-generated images
Incremental CNN-k-NN Hybrid Facial Recognition for Helmeted Facial Recognition in IoT-Enabled Smart Parking: A Case Study at Universitas Mataram
Get PDFHelmeted rider identification challenges traditional facial recognition, especially in Indonesian campuses like UNRAM, where motorbike use is prevalent and theft risks are high. This study develops a hybrid CNN-k-NN system for secure parking access. The dataset contains 2,800 augmented images (Haar Cascade crop, 224x224 grayscale), with features extracted via VGG16/ResNet and classified using k-NN (k=1, Euclidean/Cosine). The system achieves 95.62% accuracy, with precision, recall, and F1 scores of 0.96. Incremental retraining reduces processing time to under 1 second, compared to 30 minutes for full retraining. The use of cosine similarity improves accuracy slightly over Euclidean distance. This solution enhances IoT-based smart campuses by enabling efficient, real-time identification and reducing theft by improving access control. It is adaptable to low-resource environments, supporting scalable deployments in smart parking and campus security systems
LINE PATH DETECTION ON HIGHWAYS USING THE HOUGH TRANSFORM METHOD
Get PDFLane line detection on highways is an important problem in the development of intelligent transportation technology or autonomous vehicles. One commonly used method is the Hough Transform method, which is known for its excellent level of accuracy and effectiveness. Line lane detection aims to identify and monitor line lanes on highways, which helps direct and limit vehicle traffic and ensures the safety and efficiency of vehicle movement. This research uses video images from cellphone cameras that have been taken previously. The image is then processed using the Hough Transform algorithm to detect line paths on the highway. The aim of this research is to create a line lane detection system on highways that is able to identify line lanes in various road conditions by utilizing the Hough Transform Algorithm. Apart from that, it also aims to test the ability of the Hough Transform algorithm in the lane line detection system which can provide a warning if the driver is too close to the line lane, increasing safety on the road. Even though there are several obstacles such as poor road conditions, unclear or faded line paths, and busy traffic situations, the results of this research show that the Hough Transform method can be used to detect line paths on highways well, and the level of accuracy is sufficient high namely 83%
HYBRID METHOD USING NON-NEGATIVE MATRIX FACTORIZATION AND KEYWORD-BASED FILTERING FOR RECOMMENDER SYSTEM IN MOOCS
Get PDFMassive Open Online Courses (MOOCs), introduced by Dave Cormier in 2008, have revolutionized education by providing widespread access to open and participatory online learning. While MOOCs offer broad access and flexibility in learning, users often encounter challenges in selecting appropriate courses. This leads to high dropout rates. To address this issue, this research develops a recommendation system employing the Weighted Hybrid method that combines Non-Negative Matrix Factorization (NMF) and Keyword-Based Filtering (KBF). The primary objective of the research is to enhance the accuracy of course recommendations on MOOCs. The findings of this study demonstrate that the Weighted Hybrid method, integrating NMF and KBF, successfully attained a Mean Average Precision (MAP) of 0.1963. This figure signifies an improvement compared to the MAP value of 0.1855 achieved in prior research. This method effectively addresses challenges such as cold start and sparsity, while also improving scalability. Consequently, the Weighted Hybrid approach holds promise for improving the quality of recommendations, enhancing the user's learning experience, and potentially reducing dropout rates in MOOCs
Enhanced Identity Recognition Through the Development of a Convolutional Neural Network Using Indonesian Palmprints
Get PDFThe use of palmprint as an identification system has gained significant attention due to its potential in biometric authentication. However, existing models often face challenges related to computational complexity and the ability to scale with larger datasets. This research aims to develop an efficient Convolutional Neural Network (CNN) model for palmprint identity recognition, specifically tailored to address these challenges. A novel contribution of this study is the creation of an original palmprint dataset consisting of 700 images from 50 Indonesian college students, which serves as a foundation for future research in Southeast Asia. The dataset includes different scenarios with varying input sizes (32x32, 64x64, 96x96 pixels) and the number of classes (30, 40, 50) to assess the model's scalability and performance. Three CNN architectures were designed with varying layers, activation functions, and dropout strategies to capture the unique features of palmprints and improve model generalization. The results show that the best-performing model, Model 3, which incorporates dropout layers, achieved 95% accuracy, 96% precision, 95% recall, and 95% F1-score on 50 classes with 1.2 million parameters. Model 1 achieved 98% accuracy, 99% precision, 98% recall, and 98% F1-score on 40 classes with 1.7 million parameters. These findings demonstrate that the proposed CNN models not only achieve high accuracy but also maintain computational efficiency, offering promising solutions for real-time palmprint authentication systems. This research contributes to the advancement of biometric authentication systems, with significant implications for real- world applications in Southeast Asia
Geo-Sentiment Analysis of Public Opinion of X Users towards the Documentary Film Dirty Vote using the Bidirectional Long Short-Term Memory Method
Get PDFPresidential elections held every five years, often generates significant public discourse. The 2024 presidential election saw the release of the documentary Dirty Vote, which raised allegations of electoral fraud and sparked polarized opinions on social media, especially on X. This study aims to analyze public sentiment toward Dirty Vote using geo-sentiment analysis and the Bidirectional Long Short-Term Memory (Bi-LSTM) model. Data were collected from geotagged tweets, with sentiment classified as positive, negative, or neutral. The research explored various data processing techniques, including TF-IDF for feature extraction, FastText for feature expansion, and balancing methods like SMOTE and class weighting to address data imbalance. Results showed that the baseline Bi-LSTM model achieved an accuracy of 71.57% and an F1-Score of 74.05%. When enhanced with TF-IDF and FastText, accuracy increased to 77.07%, though the F1-Score dropped slightly to 72.95%. Applying SMOTE resulted in a decrease in accuracy to 76.45%, but significantly improved the F1-Score to 74.93%. Exploratory data analysis revealed that negative sentiment was most concentrated in Java Island, particularly Jakarta, and peaked during February 2024, coinciding with the documentary's release and the election period. This study significantly contributes to understanding how geographic locations influence public opinion on sensitive political issues. A lack of understanding of geographically-based sentiment patterns can hinder identifying regional needs, leading to poorly targeted policies. By integrating data analysis methods with geographical approaches, this research provides deep insights for designing more effective, data-driven public intervention strategies and supports policymaking that is more responsive to the dynamics of public opinion
ENHANCING SENTIMENT ANALYSIS OF THE 2024 INDONESIAN PRESIDENTIAL INAUGURATION ON X USING SMOTE-OPTIMIZED NAIVE BAYES CLASSIFIER
Get PDFThe inauguration of the President and Vice President of Indonesia for the 2024-2029 period has drawn significant public attention, reflecting widespread political and societal interest. This study aims to optimize sentiment analysis of public opinion on X (formerly Twitter) regarding the inauguration by enhancing the Naïve Bayes Classifier (NBC) with the Synthetic Minority Over-sampling Technique (SMOTE). Addressing the issue of class imbalance in sentiment data, the research demonstrates how SMOTE improves classification robustness. The methodology includes data crawling from X, preprocessing involving tokenization, stemming, and TF-IDF feature extraction, and sentiment labeling using TextBlob. Sentiment classification is conducted with NBC, evaluated under conditions with and without SMOTE. Metrics such as accuracy, precision, recall, and F1-score are utilized to assess performance. Results indicate that the application of SMOTE increases the accuracy of NBC from 98% to 99%, with precision improving from 0.98 to 1 and recall maintaining high levels (0.99). This 1% accuracy enhancement underscores the significance of addressing class imbalance for reliable sentiment analysis. The findings contribute to a better understanding of public sentiment during critical political events and highlight the effectiveness of SMOTE in improving text classification tasks. This research provides valuable insights into leveraging machine learning techniques for analyzing imbalanced datasets, offering implications for both academic and practical applications in sentiment analysis and political studies
Analyzing Blockchain Adoption for Copyright Certification in Lombok's Woven Industry: An Extended TAM Perspective
Get PDFThis research explores the Extended Technology Acceptance Model (TAM) and Partial Least Squares Structural Equation Modelling (PLS-SEM) to investigate the acceptability of blockchain-based digital copyright certification among traditional woven fabric SMEs (Small and Medium Enterprises) in Lombok. This research develops a blockchain-based certification system using NFTs, IPFS, and ECDSA to secure ownership, metadata, and authentication of traditional woven fabrics in Lombok. The problem addressed is the lack of understanding and acceptance of blockchain technology for copyright certification among SMEs, which can impede the protection of their innovations. The aim of this study is to analyze the variables that influence this technology's acceptance and to provide strategies for increasing its adoption. This study explores blockchain-based copyright certification adoption among Lombok's woven fabric SMEs using an Extended TAM with novel variables: Perceived Trust, Privacy, and Government Regulations. Findings from PLS-SEM reveal these, alongside traditional TAM factors, significantly impact adoption. By addressing digital literacy gaps and regulatory challenges, this research provides insights into promoting blockchain adoption through targeted training and outreach, contributing to innovation protection for traditional artisans. A quantitative method was implemented with a validated and reliable surveys distributed both online and offline to SMEs in three main woven villages in Lombok. Data analysis using PLS-SEM revealed significant impacts of perceived usefulness (PU), perceived ease of use (PEOU), Perceived Trust (PT), Government Regulations (GR), Perceived Protection (PP), attitude towards using (ATU), and behavioral intention to use (BITU) on the acceptance of blockchain technology. This study concludes that TAM factors are crucial in evaluating these SMEs' acceptance of blockchain-based copyright certification. Recommendations are provided to enhance SMEs understanding and skills in applying this technology through targeted training and outreach
Identifying Academic Excellence: Fuzzy Subtractive Clustering of Student Learning Outcomes
Get PDFEducation forms a vital foundation for a nation's future. In this digital era, while the use of Information and Communication Technology (ICT) in education is increasing, it brings increasingly complex challenges in education data management and analysis. The growing number of students each year results in a large volume of data, which would be difficult to manage if still relying on manual methods. Manual approaches are inefficient, time-consuming, prone to inconsistencies and human error, especially when identifying outstanding students in large and complex data. This research aims to implement a clustering system to group outstanding students at XYZ elementary school using the Fuzzy Subtractive Clustering (FSC) method. FSC was chosen for its ability to identify data groups based on the density of data points. FSC involves several important parameters, including radius, squash factor, acceptance ratio, and rejection ratio. Added variabel of social and spiritual values aims to enhance grouping quality by offering a broader perspective on students' character, attitudes, and social interactions. Parameter exploration shows an increase in the silhouette score from 0.20–0.45 to 0.45-0.57 and variable addition spiritual and social values, which indicates clearer cluster separation and provides better insights. The best parameters results were achieved with radius 0.3, accept ratio 0.5, reject ratio 0.04, and squash factor 1.25, resulting in a Silhouette Score of 0.57 and forming 5 student groups. Cluster results can guide special mentoring for students with low academic, spiritual, and social values, and support personalized learning programs based on each cluster’s characteristics