JOIV : International Journal on Informatics Visualization
Not a member yet
786 research outputs found
Sort by
Machine Learning Model to Predict Manganese Micronutrient Content in Oil Palm Plantation Soil Using Sentinel 1A and Sentinel 2A Image Integration
Full text linkThis study aims to predict manganese micronutrients in oil palm plantation soil using machine learning. Materials and technological tools use remote sensing with the integration of Sentinel 1A and Sentinel 2A satellites for monitoring micronutrients in peat soil in oil palm plantations. Integrating Sentinel 1A with Sentinel 2A will complement the shortcomings of Sentinel 2A, which is not free from cloud cover. Sentinel 1A has the advantage of being free from cloud cover. Meanwhile, Sentinel 2A has a high spectral resolution with 12 to 13 bands, which Sentinel 1A does not have, and only has dual polarization (VV-VH) and local incident angle (LIA). This study uses a machine learning method to obtain a model with a random forest regression algorithm and 103 soil samples in Central Kalimantan and Riau locations. The results of the model performance evaluation using integration showed MAPE and correctness of 25% and 75%, respectively. Suppose using Sentinel 1A, MAPE, and accuracy are 59.63% and 40.23%. Using Sentinel 2A, the MAPE and accuracy obtained are 48.40% and 51.59%. These results suggest that the integration of Sentinel 1A and Sentinel 2A plays a significant role, given their good predictive power. The implications of this study are the status of nutrient distribution maps, which can help determine the status of manganese micronutrients in soil in oil palm plantations for fertilizer application plans according to the needs of each oil palm plant
Applying Data Mining on Personal Computer for Document Classification
Full text linkThe typical user creates documents over many years of computer usage. As people move from computer to computer, they tend to copy the files to the new computer, because "you never know when we might need to refer to something from the past." Hence, the collection grows larger and larger, expanding to hundreds and thousands. This collection soon exceeds the ability of most people to remember what each document was, even if they have been keeping them in some order in folders – and many people fail to anticipate how the folders and subfolders should be arranged as time passes – and by the time they realize it, most find it too daunting a task to reclassify them all manually. Therefore, we sought to solve this problem using a data mining-based solution, specifically multinomial naive Bayes. We developed a document classification program to automatically categorize all documents stored on a person's personal computer hard drive, eliminating the need for manual classification. The proposed algorithm achieved a score of 0.853 for accuracy, 9,833 for precision, 0.661 for recall, and 0.767 for the F1 metric. It should be possible, with further refinement and improvement, for example by balancing the dataset and increasing its size, for this technique to be applied in practical applications that enable automatic document classifications on the computers of most computer users
Evaluation of Extreme Rainfall Occurrences Using Short-term and Long-term Standard Precipitation Index (SPI)
Full text linkThe main objective of this study is to investigate the Standard Precipitation Index (SPI), a method commonly used to determine extreme rainfall occurrences. It is also used to gauge the severity and duration of drought in meteorological studies. To highlight exceptional extreme rainfall events in selected areas, a methodology for calculating the SPI was provided in this paper using a range period and thresholds. The Standard Precipitation Index (SPI) is used to analyze monthly precipitation data from several selected rain gauge stations between 1970 and 2014. The goal of this study is to monitor the extremely moist conditions that may eventually lead to flooding. Precipitation index data from several rain gauge sites in the selected region are used to calculate the SPI time series. Additionally, SPI readings for 3 months or less may usually be used for basic drought monitoring, values for 6 months or less may be useful for monitoring agricultural impacts, and values for 12 months or more may be useful for monitoring hydrological impacts. In this study, two states affected by the monsoon season were selected: Johor and Kelantan. Two rain gauge stations were selected from these two states to calculate the SPI results. From this study, statistics on the occurrence of dry and wet events in specified areas were determined based on the SPI readings for 3-month, 9-month, 12-month, and 24-month periods. To summarize, this research demonstrates the potential of SPI to enhance our understanding of extreme rainfall events in Peninsular Malaysia
Bike Fitting System Based on Digital Image Processing on Road Bike
Full text linkThis research aims to develop a bike fitting system based on digital image processing for road bikes. The method used in this study involves using the OpenCV and MediaPipe libraries in the Python programming language to detect the rider's body pose from a video stream captured using a webcam. The body pose data is then used to calculate important angles such as elbow, hip, knee, and ankle range related to the correct riding position for road bikes. In this research, a comparison is made between the body angles obtained and the angle range considered ideal for bike fitting on road bikes. If the body angles fall within the desired range, the system will label it as "Fit”; if the body angles are outside the selected range, the system will label it as "Not Fit." The results of this study indicate that the bike fitting system based on digital image processing using a webcam can provide helpful visual feedback in improving the rider's body position for road bikes. By observing the body angles produced and seeing the "Fit" or "Not Fit" label, cyclists can adjust their position to match the ideal position in bike fitting. The system test results show a low error rate, with elbow angle having an average error of 0.81%, hip angle of 1.37%, knee angle of 0.83%, and ankle range of 1.76%. Thus, this research contributes significantly to supporting cyclists in achieving a position appropriate to their inseam height
Levenshtein Distance Algorithm in Javanese Character Translation Machine Based on Optical Character Recognition
Full text linkIndonesia has diverse art, cultures, and languages. Linguistically, Indonesia has many local languages, which makes it a diverse country, with Javanese being the regional language with the highest number of entries in the Kamus Besar Bahasa Indonesia. The Javanese script, one of the cultural symbols of Java, differs significantly from the Latin script commonly used in daily communication. In the context of cultural preservation, which is also one of the ministry's strategic steps, a translation or transfer process is needed from the Javanese script to the Latin script to the Indonesian language as an active participation in culture, with technology helping promote and introduce Indonesian culture. This study develops an algorithm-based approach to capture data images and improve translation accuracy. Transliteration is further enhanced by incorporating optical character recognition to convert character images. The study also applies a convolutional neural network (CNN) algorithm for character image recognition and a Levenshtein distance algorithm to translate Latin characters into Indonesian. The convolutional neural network (CNN) algorithm achieved an optimal % image detection accuracy of 95% at the 21st epoch. The translation process yielded a 90% word-level translation accuracy and 70% sentence-level accuracy. These results indicate that sentence translation remains suboptimal due to a lack of sufficient training data and similarities between scripts, highlighting the need for further improvements through transformer models or data augmentation
Analyzing Course Selection by MBTI Personality Types
Full text linkThis research project explores the relationship between course selection and Myers-Briggs Type Indicator (MBTI) personality types. It focuses on a private university’s IT Faculty students pursuing AI, BIA, BIO, DCN, and ST courses. In higher education, there is a limited understanding of the influence of personality types on course selection. This research aims to determine the statistically significant differences between courses with personality profiles. To achieve this, data collected from the survey is systematically analyzed to provide useful insights into the distribution of course selection among various personality types through descriptive analysis and inferential statistics tests, such as the Kruskal-Wallis Test. These assessments help examine the statistically significant difference between courses for each personality profile, supported by a p-value < 0.05. Descriptive analysis shows INFJ typically occurred in every course, showing the wide distribution of this personality type among students. Besides, the result shows INF_ types predominantly appear in median personalities across all courses among the participants. The majority of the participants have INTP personality types. The inferential statistical results show statistically significant differences in the distribution of courses for 8 MBTI personality types, while the remaining MBTI is not statistically significant. The results also show statistically significant differences between courses for each personality dimension. These results can be used to provide suggestions to students on course selection. Future research could expand this study by including a more diverse range of universities and courses and incorporating additional personality assessments
Content and Network Feature in Attention-based Neural Network for Stance Detection on COVID-19 Vaccination Tweets
Full text linkStance detection in COVID-19 vaccination utilizing tweets is crucial for several reasons, such as public health communication, monitoring vaccine sentiment, and identifying misinformation. This research aims to explore the use of attention-based neural networks for stance detection in Indonesian COVID-19 vaccination tweets. The research focuses on enhancing accuracy by integrating content and network features. The content features represent the tweet's text, while network features define the user account's following or unfollowing. The primary contribution of this research is the development of an Attention Long Short-Term Memory (AttLSTM) model for stance detection in Indonesian tweets related to the COVID-19 vaccination. This model combines content and network features to improve accuracy in classifying user attitudes. We also highlight the performance differences between Word2Vec and FastText for numerical text representation in the AttLSTM model. The research used the Indonesian COVID-19 vaccination-related tweet dataset from prior research. The dataset is extracted using user metadata to obtain content and network features necessary to represent users' interest in tweets. Our research method involves data preparation, preprocessing, extraction of content and network features, and the development of an AttLSTM model. By integrating content and network features into the AttLSTM model with Word2Vec text representation, the study demonstrates superior performance compared to the LSTM baseline model and FastText. Adding attention mechanisms to the baseline LSTM model can capture crucial information, such as the minority class inside a tweet's text. Future research will involve exploring advanced data processing methods and ensemble learning techniques to further improve the model's performance
Face Recognition for Logging in Using Deep Learning for Liveness Detection on Healthcare Kiosks
Full text linkThis study explores the enhancement of healthcare kiosks by integrating facial recognition and liveness detection technologies to address the limitations of healthcare service accessibility for a growing population. Healthcare kiosks increase efficiency, lessen the strain on conventional institutions, and promote accessibility. However, there are issues with conventional authentication methods like passwords and RFID, such as the possibility of them being lost, stolen, or hacked, which raises privacy and data security problems. Although it is more secure, face recognition is susceptible to spoofing attacks. In order to improve security, this study integrates liveness detection with face recognition. Data preparation is done using deep learning algorithms, namely FaceNet and Multi-task Cascaded Convolutional Neural Networks (MTCNN). Real-time authentication of persons is verified by the system, which provides correct identification of them. Techniques for enhancing data help the model become more accurate and robust. The system's usefulness is shown by the outcomes of the experiments. The VGG16 model outperforms alternative designs like MobileNet V2, ResNet-50, and DenseNet-121, achieving 100% accuracy in liveness detection. Face recognition and liveness detection together greatly improve security, which makes it a dependable option for real-world healthcare applications. Through the ability to differentiate between genuine and fake faces and foil spoofing efforts, facial liveness detection may boost security. This study offers insights into building biometric systems for safe and effective identity verification in the healthcare industry
Single Image Estimation Techniques for SEM Imaging System
Full text linkEstimating a single image's signal-to-noise ratio (SNR) is a critical challenge in Scanning Electron Microscopy (SEM), impacting image quality and analysis reliability. SEM images are essential for revealing structural details at the micro- or nanoscale, but noise often obscures these details, complicating interpretation. Traditional SNR estimation methods required two images to compare and assess the noise levels. SEM images are usually corrupted by noise through several operating conditions, such as dwell time, probe current, and specimen composition. This paper introduces a novel single-image SNR estimation technique, Quarsig SNR Estimation (QSE), for estimating SNR value in SEM images. This method differs from the traditional methods because it only uses a single image to obtain the SNR value without a reference image. This approach involves a single image with Gaussian noise and using the autocorrelation function (ACF) to calculate the peak value for both the original and noisy images. The peak value is the SNR value for the noisy image. QSE has outperformed the existing methods, such as Nearest Neighborhood (NN), Linear Interpolation (LI), and the combination of NN and LI by archiving the nearest SNR value to the reference measurements. This shows that QSE has significant potential for single-image SNR estimation under Gaussian noise. However, its performance under non-Gaussian noise remains a limitation. Despite this, QSE has showcased its reliability in the SEM imaging field by improving the analysis of structural details in noisy imaging conditions
Leveraging ESRGAN for High-Quality Retrieval of Low-Resolution Batik Pattern Datasets
Full text linkAs one of the world's cultural heritages in Indonesia, batik is one of the quite interesting research subjects, including in the realm of image retrieval. One of the inhibiting factors in searching for batik images relevant to the query image input by the user is the low resolution of the batik images in the dataset. This can affect the dataset's quality, which automatically also impacts the model's performance in recognizing batik motifs with complex details and textures. To address this problem, this study proposes using the Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) method to increase the resolution of batik images. By increasing the resolution, it is hoped that ESRGAN can clarify the details and textures of the initial low-resolution image so that these features can be extracted better. This study proves that ESRGAN can produce high-resolution batik images while maintaining the details of the batik motif itself. The resulting image's high PSNR and low MSE values confirm this. The implementation of ESRGAN has also been proven to improve the performance of the image retrieval system with an increase in precision and average precision values between 1-5% compared to other methods that do not implement it