Jurnal Galaksi

Accuracy Improvement of Convolutional Neural Network with Ghost Weight Normalization for Pneumonia Classification

Galih Restu Baihaqi, Shafatyra Reditha Shalsadilla, Maulida Khairunisa Argaputri — Tue, 31 Dec 2024 00:00:00 +0000

Pneumonia is a critical respiratory condition that requires accurate and timely diagnosis to ensure effective treatment. In this study, we propose the integration of Ghost Weight Normalization (GWN) into a Convolutional Neural Network (CNN) to enhance the accuracy and performance of pneumonia detection. The dataset used was derived from the Kaggle repository, comprising 5,856 chest X-ray images divided into two classes: Normal and Pneumonia. The CNN + GWN model demonstrated improved classification metrics with an accuracy, precision, recall, and F1-score of 95%, outperforming the CNN-Based model, which achieved 92%. While the CNN + GWN model required slightly longer training time and more epochs to achieve its best performance, the trade-off resulted in more robust and reliable predictions. The enhanced performance is attributed to the ability of GWN to normalize weights effectively, providing diverse normalization variations and improving training stability. These results underscore the potential of the CNN + GWN model for reliable pneumonia detection and highlight its capability to address the limitations of conventional CNN architectures.

Implementation of Recurrent Neural Network Gated Recurrent Unit (GRU) Model for Predicting Top-Tier Bitcoin

Yusuf Aliyu Adamu — Tue, 31 Dec 2024 00:00:00 +0000

Cryptocurrency investments are becoming increasingly popular due to their potential as digital assets, though high price volatility poses significant challenges for investment decision-making. This study employs the Gated Recurrent Unit (GRU) model to forecast the closing prices of five prominent cryptocurrencies: Bitcoin (BTC), Ethereum (ETH), Ripple (XRP), Binance Coin (BNB), and Dogecoin (DOGE), using historical data from Yahoo Finance spanning 2019 to 2024. The model's performance was assessed using evaluation metrics, including Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R²). The results demonstrate that BNB achieved the best performance, with a MAPE of 2.38% and RMSE of 17.03, followed by ETH and XRP, which recorded MAPEs of 2.51% and 2.64%, respectively. BTC exhibited the highest RMSE at 2280.73, highlighting its significant price volatility, while DOGE had the lowest RMSE at 0.01, despite recording the highest MAPE at 4.11%. Forecasts for the next six periods indicate that BTC and ETH are likely to experience gradual price increases, XRP and BNB are expected to stabilize, and DOGE will remain relatively stable with low volatility. The study concludes that the GRU model is effective for cryptocurrency price forecasting, but integrating it with fundamental and technical analysis could further enhance accuracy and support more informed investment decisions.

Time Facial Expression Recognition Using Optimized CNN Models for Behavioral and Emotional Analysis

Komang Diva Andi Wirawan, I Nyoman Tri Anindia Putra — Tue, 31 Dec 2024 00:00:00 +0000

Facial expression recognition is a significant field in human-computer interaction, aiming to analyze emotions such as happiness, sadness, anger, and fear. This study develops a facial expression detection system using Convolutional Neural Networks (CNN) to address challenges like lighting variations and facial angles. The research begins with collecting and preprocessing datasets, including FER-2013, to normalize, augment, and label images for seven emotion classes. The CNN model is designed with convolutional, pooling, and fully connected layers, optimized using ReLU activation, Adam optimizer, and categorical crossentropy loss function. Training is conducted on 80% of the dataset, with 20% for validation, achieving a validation accuracy of 91.7%. System performance is evaluated using precision, recall, F1-score, and real-time testing integrated with cameras, achieving an average detection accuracy of 90%. Results demonstrate the system's robustness in detecting emotions under varying conditions, highlighting its potential for applications in security, education, and emotional therapy. Future research recommends incorporating larger datasets and advanced transfer learning methods to improve system efficiency and accuracy.

Blockchain Technology: A Review Study on Improving Efficiency and Transparency in Agricultural Supply Chains

Timothy Mwewa, Gilbert Lungu, Benson Turyasingura, Yusuf Umer, Petros Chavula — Tue, 31 Dec 2024 00:00:00 +0000

The agricultural sector is vital to economic development and food security in Sub-Saharan Africa (SSA). However, persistent challenges in agricultural supply chains, such as inefficiencies, lack of transparency, and limited traceability, contribute to high post-harvest losses, unfair pricing for farmers, and reduced consumer trust. Blockchain technology, with its decentralized and transparent ledger system, offers a promising solution to these issues. This review explores blockchain’s potential to improve supply chain efficiency and transparency in SSA. By integrating blockchain with smart contracts, IoT devices, and real-time data sharing, stakeholders can enhance traceability, automate processes, and reduce transaction costs. Blockchain-based platforms provide direct market access for farmers, ensuring fair pricing and minimizing intermediary influence. Furthermore, blockchain’s immutable nature guarantees data credibility, fostering consumer trust and compliance with quality standards. Despite its potential, blockchain adoption in SSA faces challenges, including high costs, inadequate infrastructure, limited technical expertise, and low awareness of its benefits. Addressing these barriers requires affordable, scalable solutions and supportive policy frameworks. This review highlights blockchain’s transformative role in resolving inefficiencies and improving transparency in SSA’s agricultural supply chains. Collaborative efforts among governments, private stakeholders, and international organizations are crucial to fostering adoption, driving economic growth, and enhancing food security.

Towards Improved Heart Disease Detection: Evaluating Naïve Bayes and K-Nearest Neighbors in Medical Data Classification

Tue, 31 Dec 2024 00:00:00 +0000

The application of machine learning in healthcare is increasingly critical for improving diagnostic accuracy and timely treatment. This study explores the classification of heart disease using Naïve Bayes and K-Nearest Neighbors (KNN), focusing on evaluating their effectiveness through a comparative analysis. The research addresses the challenge of identifying an optimal method for heart disease classification, emphasizing the need for reliable algorithms. Using a dataset from Kaggle with detailed preprocessing, we implement Naïve Bayes and KNN to assess classification performance. The study introduces a comparative perspective on classification accuracy, precision, recall, and F1-score, revealing the strengths and limitations of each method. The results highlight the superior performance of Naïve Bayes with an accuracy of 88%, offering novel insights for data-driven healthcare decisions.

Advanced Long Short-Term Memory (LSTM) Models for Forecasting Indonesian Stock Prices

Firman Santosa, Ego Oktafanda, Hendrik Setiawan, Abdul Latif — Tue, 31 Dec 2024 00:00:00 +0000

The Indonesian stock market is a key indicator of national economic dynamics. Blue-chip stocks, including Bank Central Asia (BBCA), Bank Rakyat Indonesia (BBRI), and Bank Mandiri (BMRI), hold significant influence due to their liquidity and impact on the market index. However, their price volatility, driven by global economic conditions, monetary policies, and market sentiment, poses challenges for accurate forecasting. This study employs the Long Short-Term Memory (LSTM) model to address these challenges. LSTM, a deep learning technique, effectively handles time series data by capturing long-term dependencies and complex price patterns. Using historical stock data from 2019 to 2024, the model was trained and optimized. Evaluation metrics, including Mean Absolute Percentage Error (MAPE) and Root Mean Squared Error (RMSE), were used to assess performance. BBCA stocks achieved the best results, with a MAPE of 0.0099 and RMSE of 128.02.The findings demonstrate LSTM's robustness in forecasting stock price trends, providing investors with valuable tools for informed decision-making. This research advances predictive analytics in financial markets, particularly in emerging economies like Indonesia, and highlights LSTM’s potential to improve accuracy in volatile environments.