International Journal of Advances in Applied Sciences
Not a member yet
668 research outputs found
Sort by
Optimization of cashew apple extract as a tomato sauce substitute in chicken steak marinades
Get PDFThis study aims to optimize the use of cashew apple extract (CAE) as a substitute for tomato sauce in marinades and evaluate its effects on the chemical and sensory qualities of chicken steak. Four different marinade formulations containing varying concentrations of CAE (0, 5, 10, and 15%) were applied to chicken steak samples. Chemical analyses measured protein, fat content, and polycyclic aromatic hydrocarbon (PAH) levels, while sensory evaluations were conducted to assess tenderness, juiciness, aroma, and overall preference using a semi-trained panel. The results demonstrated that increasing CAE concentrations significantly elevated protein content (p<0.05) and reduced fat levels. PAH levels were below detectable limits in all samples, suggesting the marinade’s potential in reducing PAH formation. Sensory analysis revealed that the 5% CAE marinade was the most preferred, particularly for tenderness and juiciness. These findings suggest that CAE is a viable alternative to tomato sauce in marinades, offering both nutritional benefits and consumer acceptability
Birth data clustering to segmentation delays in birth certificate registration
Get PDFTimely and accurate birth registration is essential for ensuring access to vital public services. This study focuses on clustering birth data to identify patterns in registration delays, using data mining techniques such as the K-means algorithm. By clustering birth data from Makassar City, Indonesia, based on various demographic and birth-related criteria, the study segments the data into groups that reflect both timely and delayed registrations. The optimal number of clusters is determined using the elbow and silhouette methods. Results show that a three-cluster configuration effectively captures patterns in birth registration delays, offering critical insights for policymakers. These findings provide a foundation for improving birth registration processes, ensuring more timely registration, and guiding data-driven public policy decisions
A convolutional neural network with attention mechanism-based malaria detection from blood smear images
Get PDFWith 249 million cases and 608,000 fatalities recorded in 2022, malaria is one of the major worldwide health concerns, particularly in areas with low resources. In this paper, we propose a custom convolutional neural network (CNN) with an integrated attention mechanism to inspect malaria from blood smear images. To improve model robustness, we combined three publicly available datasets from the NIH and Kaggle. The proposed model achieved 98.20% accuracy, 97.85% precision, 98.55% recall, and 98.20% F1-score, outperforming conventional di agnostic methods. In addition, we conduct comparative analyses using two transfer learning models, ResNet50 and DenseNet. ResNet50 attained 95.06% precision, 95.44% recall, with 95.05% F1-score, while DenseNet achieved a pre cision of 87.96%, recall of 88.33%, and F1-score of 87.90%. For interpretability, Grad-CAMandsaliency map visualizations highlighted key image regions, with saliency maps offering finer pixel-level localization. These results highlight the potential of our attention-based CNN as a feasible, interpretable diagnostic tool for malaria, particularly in low-resource settings
Analysis of feature reduction for identifying stress levels electroencephalogram signal based
Get PDFStress identification based on electroencephalogram (EEG) signals has become a rapidly growing research topic, with the main approaches utilizing features from the frequency domain and time-frequency domain. This research aims to combine principal component analysis (PCA) and independent component analysis (ICA) for feature extraction to improve the accuracy of stress identification. Additionally, PCA+ICA features are reduced from 64 to 32 columns to optimize computational efficiency without losing important information from the EEG signal. The stress identification models used in this research include Ensemble, naive Bayes, and support vector machine (SVM). The data used are from the SAM-40 task Stroop color trials 1, 2, and 3. Experimental results indicate that the combination of PCA+ICA features improves accuracy only in the ensemble method. Reducing PCA+ICA features from 64 to 32 columns led to an improvement in accuracy only for Stroop trial 2 data with the naive Bayes method
Artificial intelligence-based multi-key security for protected and transparent medical cloud storage
Get PDFEnsuring the security and privacy for the patient medical records and medical reports data is a crucial challenge as cloud-based healthcare technologies become more prevalent. For cloud-hosted medical data, internet of things (IoT) and artificial intelligence (AI) technologies shows best solutions for the challenges in the medical domain. This study suggests a Secure and Transparent Multi-Key Authentication Framework that makes use of AI. Using Z-score normalization, the framework first preprocesses the data before clustering to create a multi-level multi-key security structure. The physics-informed triangulation aggregation neural network (PITANN) model in the study reduces computation costs by minimizing overhead, ensuring secure handling of location-based and medical data for enhanced data classification and encryption effectiveness. A multi-key derivation of an elliptic curve, the ElGamal cryptography scheme is presented, which allows for safe multi-key encryption with little increase in the length of the ciphertext. This method guarantees safe, confidential access to cloud-hosted encrypted health information. An envisioned amalgamation improves flexibility by enhancing performance metrics such as speed of computation while safeguarding patient information through enhanced security measures and ensuring precise medical record integrity within virtual healthcare systems
Prediction index drought use neural network based rainfall
Get PDFProlonged dry seasons compared to rainy seasons often lead to drought, making drought index observations essential. In Indonesia, drought monitoring commonly uses the standardized precipitation index (SPI), yet there is no common standard for drought index measurement. Therefore, this research applies the Z-score index (ZSI) and China-Z index (CZI), which, like SPI, are rainfall-based drought indices but have rarely been explored in previous research. To predict ZSI and CZI, this research compares the weighted moving average (WMA) and multilayer perceptron (MLP) methods. Two input scenarios are tested: the previous two periods (t-2, t-1) and the previous three periods (t-3, t-2, t-1). The results show that MLP outperforms WMA, with the best performance achieved by the MLP model at a mean absolute percentage error (MAPE) of 4.177% using the three variable input scenario and MLP architecture 3-6-10-1
A course review analysis using bidirectional long short-term memory model
Get PDFIn recent years, sentiment analysis and online review analysis have gained popularity as critical components in the growth and development of educational courses. An innovative method has been created to increase the quality of learning experiences by rapidly collecting relevant data from course comments. This technique leverages bidirectional encoder representation from transformers (BERT) for word vector training. When combined with a learning mechanism, the recommended BERT accurately predicts the sentiment of online course reviews. Additionally, a dual-channel model based on Bi-directional long short-term memory (Bi-LSTM) is employed to improve sentiment data and semantics. Following data collection from the Coursera dataset, preprocessing approaches such as tokenization, stop words removal and sentence metric creation are applied to convert input data into word vectors and identify fundamental text units using text segmentation. The results demonstrate the proposed approach’s superiority over existing methods, offering an accuracy of 81.45%, recall of 94.9%, precision of 93.7%, and F-score of 93.7%
A hybrid features based malevolent domain detection in cyberspace using machine learning
Get PDFThe rise of social media has changed modern communication, placing information at our fingertips. While these developments have made our lives easier, they have also increased cybercrime. Cyberspace has become a refuge for modern cybercriminals to conduct destructive actions. Most cyberattacks are carried out through malicious links shared on social media platforms, emails, or messaging services. These attacks can have serious consequences for individuals and organizations, including financial losses, sensitive data breaches, and damage to reputation. Early identification and blocking of such links are crucial to protecting internet users and securing cyberspace. Current research uses machine learning (ML) algorithms to detect malicious hyperlinks based on observed patterns in uniform resource locators (URLs) or web content. However, cyberattack tactics are constantly changing. To address this challenge, this paper introduces a robust method that performs a fine-grained analysis of URLs for classification. Lexical and n-gram features are examined separately, with URL n-grams represented using Word2Vec embeddings. The results from hybrid feature sets are combined using a logistic regression (LR) model to increase overall classification accuracy. This robust method allows the system to use both the structural components of the URL and the fine-grained patterns obtained by the n-grams
An innovative design and development of multilevel inverter for a wind energy conversion system
Get PDFThe drawbacks of fossil fuel-based energy sources, including high costs, pollution, scarcity, and environmental damage, highlight how urgent it is to switch to renewable energy sources. Multilevel inverters (MLIs) are currently required for the production of wind electricity. In this research, to get a reduced harmonic distortion, use 31-level inverter based on shifted carrier-pulse width modulation (SC-PWM) is developed for wind power generation using MATLAB/Simulink. It aids in minimizing the total harmonic distortion (THD) to 3.20, and the output voltage is enhanced by the suggested MLI. Wind energy extraction is optimized by combining with a proportional integral derivative (PID) control system. MATLAB/Simulink has been used to make sure the MLI structure and look into the implementation of wind energy conversion systems using a permanent magnet synchronous generator (PMSG). In order to show that the suggested inverter architecture improves power conversion efficiency and stability in renewable energy systems, the study also examines power efficiency, system dependability, and the viability of large-scale applications. Additionally, the study investigates grid integration, modulation strategies, and switching losses to guarantee increased sustainability, dependability, and efficiency in wind energy applications while lowering operating costs
Earthquake epicenter prediction from the Java-Bali radon gas telemonitoring station using machine learning
Get PDFPredicting the location of earthquake epicenters is a critical aspect of earthquake forecasting, as it complements efforts to determine the time and magnitude of seismic events. This research addresses the challenge posed by the uncertainty in epicenter locations, particularly along the extensive plate faults of Indo-Australia and Eurasia. In these regions, effective earthquake prediction is compromised without accurate epicenter information, impeding mitigation strategies and complicating disaster impact estimation. The primary objective of this study is to devise an algorithm for forecasting earthquake epicenter locations by harnessing variations in radon gas concentrations on southern Java Island, Indonesia, as a predictive precursor. Using a supervised machine learning approach, this study integrates radon gas concentration data to predict the distance between a radon gas telemonitoring station and the impending earthquake epicenter. Three distinct machine learning algorithms were evaluated using data from six Java-Bali radon gas telemonitoring stations within an early warning system. The random forest algorithm emerged as the most effective, yielding an average root mean square error of 453.10 kilometers. The findings of this research significantly contribute to earthquake risk mitigation efforts. This work enhances our capability to anticipate seismic events, and more effective disaster preparedness and response strategies in earthquake-prone regions