Inform : Jurnal Ilmiah Bidang Teknologi Informasi dan Komunikasi

Chili Price Prediction One Year Ahead Using the Gated Recurrent Unit Method

2025-03-24T13:52:14+07:00

Chili is an important commodity in the Indonesian economy, and it experiences significant price fluctuations. This is due to several factors, such as demand and supply in the market and climate and temperature factors. The prediction of chili prices is important to minimize the risk of loss to the community and the government. Previous studies have used ARIMA, SVR, MLP, and RNN methods to predict chili prices. However, these methods are still considered less effective, especially in processing large amounts of information. In this study, the Gated Recurrent Unit (GRU) method is used as an alternative, and this method is designed to overcome obstacles to remembering long information. The results show that GRU provides excellent prediction performance, with a MAPE value below 10% for all types of chili peppers. The test results that have been carried out show that the size of the window data affects the accuracy of predicting chili prices. In addition, the combination of hyperparameters, namely hidden neurons and epochs, also influences the accuracy of predicting chili prices. The lowest MAPE value is obtained in predicting large red chili pepper prices of 4.850% with 512 hidden neurons and 100 epochs. Then, a MAPE of 6.434% was obtained using 512 hidden neurons and 100 epochs to predict curly red chili pepper prices. In predicting the price of green bird's eye pepper, a MAPE of 7.288% is obtained with a combination of 512 hidden neurons and 150 epochs. Finally, the lowest MAPE value is obtained in predicting the price of red bird's eye pepper at 6.452% with a combination of 512 hidden neurons and 150 epochs.

Identification of Social Media Addiction Levels on "TikTok" Among Students Using Mamdani Fuzzy Logic

2025-03-24T13:52:02+07:00

According to data from the Global Web Index in their early 2023 survey, advertising sources published by Byte Dance indicate that TikTok has 109.9 million users aged 18 and above in Indonesia alone. This highlights the inevitable trend of teenagers becoming addicted to social media, particularly the TikTok app. Related parties are starting to notice this addiction issue, along with a noticeable decline in academic performance, possibly as a result of students neglecting their time by spending it on social media (TikTok). Analyzing variables such as usage duration, productivity impact, and neglect of social interactions aims to measure TikTok addiction among students using the Mamdani fuzzy logic method. Of the 75 respondents surveyed, 17.33% were identified as "Not Addicted", 72% were identified as "Addicted", and the remaining 10.67% were identified as "Highly Addicted". The results show that most students are already identified as addicted to TikTok. The Mamdani fuzzy logic method is considered effective in handling uncertainties in human behaviour. By utilizing MATLAB’s Fuzzy Logic Toolbox, this research models and evaluates addiction levels, providing insights to reduce excessive TikTok usage. This study also provides a strong foundation for future research exploring social media addiction on other platforms or among different populations. It encourages using technologies like MATLAB’s Fuzzy Logic Toolbox in educational contexts to address other complex problems.

Enhanced AI-Based Navigation System for The Visually Impaired

2025-01-16T11:05:49+07:00

This work introduces an Artificial intelligence (AI) based navigation system using the Raspberry Pi single-board computer. Its primary aim is to assist visually impaired users with precise navigation, achieved through machine learning algorithms for object detection. The development follows an agile methodology, emphasizing flexibility. The work explores integrating essential technologies into a system divided into two major subsystems: hardware and software. The hardware subsystem consists of a Raspberry Pi processor, a camera, an ultrasonic sensor, and a power source, collecting data on road conditions and traffic. The software employs TensorFlow and OpenCV to process data and provide optimized routes. The processed images were classified and identified using the YOLOv3 algorithm. The ultrasonic sensor could measure object distance with about 99.8% accuracy correctly. The test results demonstrate that the AI-based navigation system enhances user experiences and interaction with their environment by simplifying transportation and delivering accurate routes. It effectively analyzes and processes data obtained from the environment, improving accessibility for visually impaired individuals. The work concludes by discussing potential applications and future directions for AI-based navigation systems. It highlights the importance of affordable and accessible technology in improving transportation infrastructure, showcasing the potential for low-cost technology to enhance accessibility and mobility.

Analysis of Driver Drowsiness Detection System Based on Landmarks and MediaPipe

2025-01-16T11:05:42+07:00

Driver drowsiness is one of the leading causes of traffic accidents, especially during long-distance journeys. This study developed a detection system based on landmarks and the MediaPipe framework to analyze drowsiness through eye blink duration. The system employs coordinate point initialization using regression trees to accurately detect objects, such as eyes. The research data consists of 30 videos, each lasting 30 seconds, collected from four Trans Java bus drivers. The videos were extracted to identify facial detection histograms and analyzed based on eye blink duration. The testing results showed a detection accuracy of 81% with an error rate of 19% for distances of 10 to 100 cm, while testing with 30 videos achieved an average accuracy of 88.745% and a Mean Squared Error (MSE) of 7.615%. The test results show that CNN outperforms MediaPipe in detecting drowsiness, with a higher average accuracy of 76.79% compared to 73.83% and a lower MSE value of 47.33 compared to 48.27. CNN is also more consistent in handling extreme lighting variations, while MediaPipe excels in processing efficiency, making it suitable for devices with limited resources. This study demonstrates that the landmarks and MediaPipe-based system effectively and innovatively detects drowsiness, offering a solution to improve driver safety during trips.

Comparative Analysis of Decision Tree and Artificial Neural Network Methods for Predicting Potential Heart Disease

2025-03-24T13:51:43+07:00

Prediction models have been used in various fields such as health, education, and industry. This system can connect various data collected to be used as learning for the system in solving a problem similar to the data used as learning. The prediction model involves various elements such as mathematics, machine learning, and statistics. Heart disease remains a leading cause of mortality globally, and accurate prediction models are crucial for early detection and treatment. However, existing models often struggle with dataset imbalance, leading to suboptimal performance. This study aims to compare the performance of Decision Tree and Artificial Neural Network (ANN) models, including the Elman and Jordan variants, to identify the most suitable prediction model for heart disease with a quantitative study. The type of ANN used is multi-layer with the Elman and Jordan models. However, a comparative analysis of heart disease objects was carried out using the K-Nearest Neighbor (KNN) and Naïve Bayes methods, which resulted in Naïve Bayes being better than KNN. From all the processes that have been carried out, the researchers obtained results from precision, recall, and F1-score, which were classified as poor, with an average of 55%. The Decision Tree model achieved an average accuracy of 79%, while the Elman and Jordan networks achieved 87% and 86%, respectively. However, precision, recall, and F1-scores were relatively low, averaging 55%, likely due to dataset imbalance. The accuracy results obtained are also not always directly proportional to the amount of data used. There is a significant decline at the beginning of the process, but the accuracy obtained continues to increase until all the data is used. Apart from that, there was a spike in precision, up to 80%, in several implementation processes with prediction models. Based on the results obtained in the implementation process, it can be said that the Elman Network is superior to other methods when using accuracy benchmarks. However, the relatively low precision, recall, and F1-score results indicate the model's performance is lacking.

Optimizing Face Recognition and Emotion Detection in Student Identification Using FaceNet and YOLOv8 Models

2025-03-24T13:51:34+07:00

The growing need for efficient student identification systems has driven advancements in face recognition and emotion detection technologies. This research presents a single-photo-based system that integrates YOLOv8 for face detection and FaceNet for generating unique facial embeddings, ensuring high-precision student identification and consistent emotion detection under diverse conditions. YOLOv8 localizes faces within images, while FaceNet processes them to generate embeddings for recognition. Emotion detection is performed using these embeddings or an auxiliary emotion classification model. The methodology includes pre-processing images into 64x64 grayscale format, employing image augmentation to enhance model generalization, and evaluating performance using accuracy, precision, recall, and F1 score metrics. The experimental dataset comprises 10 formal student photos, with testing conducted on 100 images. Results demonstrate 94% accuracy in face recognition with augmentation, surpassing 92% without it. Emotion detection achieves 95% accuracy in identifying seven emotions, including angry, happy, sad, neutral, fear, disgust, and surprise, despite variations in expression, lighting, and angle. This system provides a scalable and efficient solution for educational applications such as automated student identification, attendance monitoring, and emotion-based learning management. Its potential spans short-term automation in attendance and mental health monitoring, medium-term improvements in personalized learning and campus security, and long-term AI-driven educational advancements while addressing privacy and social acceptance challenges.

Skin Lesion Classification Using YOLOv11 on the HAM10000 Dataset

2025-03-24T13:51:26+07:00

Skin cancer represents a significant global health concern due to its high mortality rate. Early and accurate detection is crucial but often hindered by the limitations of traditional diagnostic methods. This research applies the YOLOv11 algorithm for skin lesion classification directly from dermoscopic images using the HAM10000 dataset (10,015 images, 7 skin lesion classes). The primary objectives are to evaluate YOLOv11's performance in multi-class classification and assess the impact of data augmentation (rotation, horizontal flipping) in addressing class imbalance. The methodology involved two experiments: training YOLOv11 on the original and augmented datasets and comparing its performance with multi-stage architectures (VGG19 and ResNet50). Five pre-trained YOLOv11 models were tested using accuracy, precision, recall, and F1-score metrics. Results showed the YOLOv11x-cls model trained on the augmented dataset achieved the best performance among YOLOv11 models (accuracy 84.74%, precision 83.94%, recall 84.74%, F1-score 84.06%). However, VGG19 recorded the highest accuracy (89.68%). Data augmentation effectively improved model performance by mitigating class imbalance. This study also indicates that multi-stage architectures perform better in skin lesion classification tasks than single-stage architectures. The key contributions of this research are: (1) a comprehensive performance comparison of YOLOv11 with VGG19 and ResNet50 for skin lesion classification and (2) empirical validation of data augmentation's effectiveness in improving model performance. This study demonstrates that YOLOv11 can achieve competitive performance in skin lesion classification despite not surpassing the performance of multi-stage architectures.

Design of Network Device Placement and Bandwidth Allocation Using Simple Queue Method

2025-03-24T13:51:19+07:00

Network usage that is not properly managed often leads to unfair bandwidth distribution and unstable internet connections. This problem causes slow internet connections and disruptions to activities that require internet connectivity. The solution to this problem is structured bandwidth management using Simple Queue. Network configuration is performed on the router to distribute bandwidth to all ports connected to the switch, and then PCQ (Per Connection Queue) can limit bandwidth for each user. The research methodology uses the Network Development Life Cycle (NDLC) with analysis, design, simulation, implementation, monitoring, and management phases. However, this research only covers the simulation phase and does not include the implementation, monitoring, or management stages. This research aims to design network topology and optimize the network, ensuring even and efficient bandwidth distribution to support all activities using internet connections properly. The testing results indicate that the download and upload speeds closely approach the bandwidth allocation set using the PCQ method, which is 10 Mbps, demonstrating optimal network performance. Bandwidth management using Simple Queue proves effective in managing and optimizing bandwidth usage in complex environments, providing fair bandwidth allocation to all users in the network. This reduces the possibility of network congestion and ensures all activities requiring an internet connection can run optimally.

New Approach to The Perceptron Algorithm with Quantum Computing for Prediction Analysis of Rice Imports in Indonesia

2025-03-24T13:51:11+07:00

Rice imports are crucial to ensure a country's food availability, especially when domestic production is insufficient. Because rice is the staple food of Indonesians, a spike in rice prices could cause social unrest. Rice imports have a strategic role in maintaining food stability and reducing the risk of price instability. This research aims to utilize the Quantum Perceptron algorithm to predict rice imports more effectively. Quantum Perceptron is a new approach that combines the principles of quantum mechanics with artificial intelligence to improve prediction performance. Researchers used data on the number of rice imports from the leading countries of origin obtained from the Central Statistics Agency using 7 variables x1 to x7. The research results show that the quantum perceptron algorithm can make predictions very well, proven by a perfect accuracy of 100% with a total of 20 epochs. This result is still better than the classical perceptron, which has 100% accuracy but with a larger number of epochs, namely 50. Quantum perceptron has better performance and shorter time, which can be seen from the smaller number of epochs compared to the classical perceptron.

Leveraging Self-Organizing Maps for Effective Image Restoration

2025-03-24T13:51:05+07:00

Computer vision relies critically on image restoration techniques to recreate clear images from damaged observations. Traditional methods face challenges when attempting to balance removing image noise and protecting image details. A novel framework that employs Self-Organizing Maps (SOMs) establishes a practical approach to restoring images will be investigated in this paper. Our restoration approach starts with image pre-processing, which feeds trained SOM features into a deep neural network to optimize outcome quality. This research evaluates our approach on benchmark datasets, achieving quantitative results: Our SOM-based method produces restoration outcomes with an average Peak Signal-to-Noise Ratio (PSNR) performance of 32.10 dB and Structural Similarity Index (SSIM) values of 0.894 that exceed state-of-the-art GAN-based restoration (31.75 dB, 0.890). According to this research, UTH can restore images by achieving enhanced clarity with preserved details. The successful merger between SOMs and deep learning architectures is our study's distinctive feature while creating opportunities for additional image processing applications.

Entity Extraction and Annotation for Job Title and Job Descriptions Using Bert-Based Model

2025-03-24T13:50:58+07:00

This research paper investigates Named Entity Recognition (NER) within Indonesia’s job vacancy domain, employing state-of-the-art Bert-based models. The study presents a detailed data collection and preprocessing methodology, followed by the Bert-based model’s fine-tuning for enhanced NER. The dataset comprises 48,673 job vacancies collected from the JobStreet website in July 2023, specifically focusing on multi-entity recognition, including job titles and job descriptions. An original annotation algorithm was developed using Python and Laravel for precise entity recognition. In addition, this paper provides an extensive literature review of NER and Bert-based models and discusses their relevance in the context of the Indonesian job market. The outcomes highlight the efficacy of our BERT-based model, attaining an average accuracy of 78.5%, a precision of 79.7%, a recall of 81.1%, and an F1 score of 80.8% in the Named Entity Recognition (NER) task. The study concludes by discussing the implications, limitations, and future directions, underscoring the model’s potential applicability in streamlining job matching and recruitment processes in Indonesia and beyond. This research contributes to the field by providing a robust framework for NER in job vacancies, highlighting the potential for improved job matching, and proposing enhancements for future model development and application in other languages and regions.

Comparative Analysis of PCOS Classification Using Random Forest: Integration of Mutual Information, SMOTE-Tomek, and Outlier Handling

2025-03-24T13:50:51+07:00

Polycystic Ovary Syndrome (PCOS) is a hormonal disorder affecting women of reproductive age, with a global prevalence rate of 8–13%. However, approximately 70% of cases remain undiagnosed. This study aimed to develop and compare eight Random Forest classification models for PCOS detection using a publicly available Kaggle dataset. The methodology incorporated three key preprocessing techniques: outlier handling using the Interquartile Range (IQR) method, feature selection through Mutual Information, and class imbalance via SMOTE-Tomek. The results revealed that the best-performing model, which applied outlier removal and SMOTE without feature selection, achieved an accuracy of 94.11%. This result significantly outperformed the baseline Random Forest model, which achieved an accuracy of 87.27% without the application of any preprocessing techniques, such as outlier removal, SMOTE, or feature selection. Moreover, the model utilizing only SMOTE for class balancing achieved an accuracy of 93.84%, underscoring the importance of addressing class imbalance in enhancing classification performance. Notably, feature selection did not consistently improve accuracy, as Random Forest inherently handles feature redundancy, capturing complex feature interactions. These findings highlight the importance of tailored preprocessing strategies, particularly outlier handling and class balancing, for optimizing medical data classification. Future research should explore clinically informed feature selection techniques and assess the generalizability of these findings across diverse datasets to enhance the clinical relevance of PCOS detection models.

Design and Analysis of Microstrip Antenna Using Double Slot Patch At 28 GHz Frequency for 5G Technology

2025-03-24T13:50:43+07:00

The advancement of telecommunication technology has increased the number of wirelessly connected devices, leading to a surge in data traffic. Fifth-generation (5G) technology utilizes millimeter waves at a 28 GHz frequency to support extensive connectivity and high data transmission rates. However, these high frequencies face challenges, such as limited signal range. This study aims to design and analyze microstrip antennas with 8×8 and 16×16 configurations using a dual-slot method (T-slot and U-slot) to enhance antenna performance, particularly bandwidth and gain. The design approach involved simulations using CST Suite Studio software. The simulation results show that the 8×8 antenna achieved a return loss of -29.40 dB, a VSWR of 1.07, a bandwidth of 1.4 GHz, and a gain of 15.6 dBi. Meanwhile, the 16×16 antenna achieved a return loss of -35.95 dB, a VSWR of 1.03, a bandwidth of 1.44 GHz, and a gain of 18.6 dBi. These results demonstrate that increasing the number of antenna elements and applying dual-slot techniques significantly improves performance, making it a potential solution for 5G communication systems requiring stronger signals and wider coverage.