Indonesian Journal of Electrical Engineering and Informatics (IJEEI)
Not a member yet
776 research outputs found
Sort by
Intelligent Bankruptcy Prediction Models Involving Corporate Governance Indicators, Financial Ratios and SMOTE
Get PDFThis study enhances bankruptcy prediction models by investigating synergies between predictors, utilizing a diverse dataset of financial statements and corporate governance data. Rigorous feature selection identifies key financial ratios (FRs) and corporate governance indicators (CGIs) to enhance model interpretability. Multiple machine learning algorithms construct and assess the models, including Logistic Regression, Decision Trees, Random Forests, Support Vector Machines, and Neural Networks. Integration of CGIs with FRs aims to identify effective combinations that improve model performance with an accuracy respectively 90%, 95%, 97%, and 98%. Researchers explore feature weighting techniques and ensemble methods, examining their impact on accuracy, sensitivity, and specificity. The study also explores how regulatory frameworks and governance practices affect bankruptcy prediction, analyzing data across periods to uncover changes in predictive power under varying conditions. The findings have implications for investors, institutions, and policymakers, offering more accurate risk assessments and emphasizing the interplay between financial performance and governance quality for corporate well-being
Enhancing Accuracy for Classification Using the CNN Model and Hyperparameter Optimization Algorithm
Get PDFThe Convolutional Neural Network (CNN) is a widely employed deep learning model, particularly effective for image recognition and classification tasks. The performance of a CNN is influenced not only by its architecture but also critically by its hyperparameters. Consequently, optimizing hyperparameters is essential for improving CNN model performance. In this study, the authors propose leveraging optimization algorithms such as Random Search, Bayesian Optimization with Gaussian Processes, and Bayesian Optimization with Treestructured Parzen Estimators to fine-tune the hyperparameters of the CNN model. The performance of the optimized CNN is compared with traditional machine learning models, including Random Forest (RF), Support Vector Classification (SVC), and K-Nearest Neighbors (KNN). Both the MNIST and Olivetti Faces datasets are utilized in this research. In the training procedure, on the MNIST dataset, the CNN model achieved a minimum accuracy of 97.85%, surpassing traditional models, which had a maximum accuracy of 97.50% across all optimization techniques. Similarly, on the Olivetti Faces dataset, the CNN achieved a minimum accuracy of 94.96%, while traditional models achieved a maximum accuracy of 94.00%. In the training-testing procedure, the CNN demonstrated impressive results, achieving accuracy rates exceeding 99.31% on the MNIST dataset and over 98.63% on the Olivetti Faces dataset, significantly outperforming traditional models, whose maximum values were 98.69% and 97.50%, respectively. Furthermore, the study compares the performance of the CNN model with three optimization algorithms. The results show that integrating CNN with these optimization techniques significantly improves prediction accuracy compared to traditional models
Fractal Analysis of Time Domain Dielectric Response to Reduce Complexity of Insulation Condition Diagnosis Methodology
Get PDFThe health of cellulosic insulation, present in a power transformer, continuously degrades due to its exposure to paper moisture and high temperature. The moisture content of such insulation further accelerates the ageing phenomena. Recent developments made in the field of power transformer insulation diagnosis show that conditionbased maintenance of power transformers is more important rather than time-based maintenance. On the other hand, utilities always prefer to monitor the condition of power transformers in short measurement time. The present work proposes a fractal analysisbased condition monitoring technique. The method utilizes only a 600 s measured profile of polarization current. This paper estimates various ageing-sensitive performance parameters evaluated from fractal features for insulation diagnosis. The suggested technique can be used in a non-intrusive way to estimate performance measures such as %pm and paper conductivity. With the least amount of shutdown time, this technique quickly assesses the insulating state of power transformers. This strategy has shown to be more successful than existing approaches for monitoring insulation status
Bengali Word Detection from Lip Movements Using Mask RCNN and Generalized Linear Model
Get PDFSpeech processing with the help of lip detection and lip reading is an advancing field. For this, we need proper algorithms and techniques to detect lips and movements of lips perfectly. Lip detection and configuration are the most important parts of speech recognition. In this paper, we focus on detecting the lip segment properly. Mask R-CNN (Regional Convolutional Neural Network) performs object detection and instance segmentation per video frame to detect the lip segment. The process of mask R-CNN adds only a small overhead to Faster R-CNN and is quite simple to train, running at 5 frames per second. The Mask R-CNN involves keypoint detection which helps to extract the location of the lip landmarks pixel by pixel. Once the lip region is extracted and the landmarks are highlighted, we observe how the lip landmarks change as the object's lips move over time to each Bengali word. The keypoint changes that are observed during each millisecond are then the landmarks used to train the GLM (Generalized Linear Model). In addition, we compare the performance of GLM with Naive Bayes, Logistic Regression, and Decision Tree. The GLM has exhibited the highest 91.8% accuracy, whereas the Naive Bayes, Logistic Regression, and Decision Tree show the accuracy of 87.1%, 38.3%, and 82.2%, respectively
Novel Polar Coded MIMO Power Domain NOMA Scheme for 5G New Radio (NR)
Get PDFThe use of Polar coded Multiple Input Multiple Output Power Domain Non-Orthogonal Multiple Access (MIMO PD-NOMA) technology has the potential to greatly improve the capacity and spectral efficiency of 5G NR systems. From the on-going research, there is a combination of polar coded NOMA and Polar coded MIMO techniques are approached separately with other channel coding techniques. This paper introduces a novel approach to combine polar coded with MIMO power domain NOMA to enhance the system performance. MIMO Power Domain NOMA that utilizes polar codes for channel coding and power allocation. By combining the benefits of NOMA and MIMO, which permits multiple users to share frequency-time resources simultaneously and the MIMO employs multiple antennas to increase diversity gain and spatial multiplexing gain. The proposed scheme provides effective utilization of radio resources where the polar codes are an optimal choice for 5G NR systems due to their strong error correction capability and low complexity decoding. Successive Cancellation List -Singular Value Decomposition adaptive scaling algorithm (SCL-SVD) is proposed in the polar decoding process. The suggested method attains 6.5 dB coding gain and improved throughput of 80.34% using MATLAB simulation. The proposed model compared with the other existing model such as Power Domain NOMA (PD-NOMA), multiple input single output NOMA (miso-NOMA) and multiple input multiple output NOMA (mimo-NOMA) in terms of Bit Error Rate (BER) and Signal to Noise Ratio (SNR). This scheme has the potential for practical implementation and can play a crucial role in meeting the increasing demands of future wireless communication systems
Assessing MANET Routing Protocols: Comparative Analysis of Proactive and Reactive Approaches with NS3
Get PDFMANETs are dynamic, decentralized networks that employ mobile nodes as hosts and routers. High mobility and frequent topology modifications make routing triky in these networks. This research compares the efficacy of four MANET protocols: DSR, AODV, OLSR, and DSDV. We evaluate various protocols utilizing the NS3 simulator for PDR, throughput, control overhead, and delay. We analyze each protocols strong points and weaknesses under varied node densities, pause durations, and network sizes. DSR has the greatest PDR and lowest control overhead, making it ideal for dynamic networks. OLSR maintains high throughput and short delay despite increasing control overhead. DSDV has the maximum throughput but significant control overhead and PDR in bigger networks. AODV performs well in smaller networks but degrades significantly as network size rises. This research illuminates MANET routing protocol trade-offs, helping to build more resilient and efficient communication techniques for diverse application situations. Our results imply that DSR is best for dynamic contexts and OLSR for route availability and low latency
Enhancing Confidence In Brain Tumor Classification Models With Grad-CAM And Grad-CAM++
Get PDFBrain tumors are a terrible and dangerous health problem, often posing a significant threat to individuals due to their high probability of death. Detecting these tumors at an early stage is crucial, as it not only increases the chances of successful treatment but also plays a pivotal role in reducing total healthcare costs. Early detection allows medical professionals to take action quickly, enabling a more targeted and effective treatment approach. Numerous studies are currently employing Machine Learning (ML) and Deep Learning (DL) to classify brain tumors, promising improved accuracy and efficiency in tumor identification for potential breakthroughs in medical diagnosis. However, a significant challenge lies in these models being "black box" as their complex inner workings are not easily understood by humans. Explainable Artificial Intelligence (XAI) refers to the capability of an artificial intelligence (AI) system to provide understandable and interpretable explanations for its decisions or predictions. In this study, we propose a classification model based on various network architectures, namely DenseNet201, DenseNet169, Xception, MobileNetV2 and ResNet50. We then used Grad-CAM and Grad-CAM++ to interpret the model's results, evaluating its ability to distinguish important features in Magnetic resonance imaging (MRI) images of brain tumors during the decision-making process. The integration of Grad-CAM and Grad-CAM++ enhances the interpretability of the brain tumor classification model, providing valuable evidence of its effectiveness by focusing on crucial features in MRI images of brain tumors during decision-making. Research results contribute to the development of systems that support early diagnosis of tumors. This contribution is pivotal as it not only enhances the model's transparency but also validates its effectiveness in accurately identifying brain tumors
Collaborative Online International Learning to Address Mental Health Across Cultures with an Islamic Perspective
Get PDFThis study examines the impact of a Collaborative Online International Learning (COIL) project on enhancing students’ cross-cultural understanding, collaborative skills, and problem-solving abilities in addressing mental health issues through technology, enriched by insights from an Islamic perspective. The project connected students from the International Islamic University Malaysia’s (IIUM) Operating Systems course and Shenandoah University’s Occupational Therapy in Mental Health Practice course, fostering a dynamic, interdisciplinary learning environment. Students worked in diverse teams, engaging in activities such as video introductions, infographic creation, and presentations on technological applications in mental health, facilitated by platforms like Zoom, Google Sites, and WhatsApp. Evaluations, including Programme Outcome (PO) analyses, revealed that over 80% of students achieved “Acceptable” or higher levels in applying engineering knowledge (PO1) and problem analysis (PO2), reflecting the success of the project in meeting its learning objectives. Student reflections captured on Flipgrid further underscored the project’s impact, with participants highlighting improved cultural sensitivity, adaptability to a global professional context, and collaborative problem-solving despite challenges such as time zone differences. The inclusion of Islamic perspectives provided a holistic lens, emphasizing spiritual and technological solutions to mental health issues through values such as patience (sabr), gratitude (shukr), and trust in Allah (tawakkul). This study underscores COIL’s potential as a transformative pedagogical approach for preparing students to navigate multicultural, technology-driven environments while fostering global mental health awareness. It offers actionable insights for educators and policymakers seeking to integrate cultural and religious perspectives into interdisciplinary educatio
Analyzing Success Factors in Developing Mobile Applications for Farmers in Thailand Using Analytic Hierarchy Process Technique
Get PDFThis study applies the Analytic Hierarchy Process (AHP) as a robust decision-making framework to address critical gaps in risk management and criteria prioritization within mobile application development for the agricultural sector, specifically focusing on Thai rice farmers. The research identifies essential factors influencing technology adoption through input collected from 100 rice farmers in Surin Province. Using AHP, these factors were systematically ranked, with "ease of use," "provision of up-to-date information," and "support" emerging as the most significant criteria. Based on these insights, three mobile application prototypes were developed, with Mobile App 1 achieving the highest AHP score of 0.633, demonstrating superior alignment with user requirements. Subsequent evaluations of user satisfaction reinforced these findings, with "ease of use" scoring the highest (4.60), followed by "perceived usefulness" (4.10). The findings underscore AHP’s efficacy in mitigating risks and aligning application features with user demands, thereby enhancing adoption effectiveness. This study contributes novel insights into leveraging AHP as a precision tool for guiding mobile application development in agriculture and provides a replicable framework for addressing user-centric challenges. Future research should investigate integrating AHP with emerging technologies to drive innovation and sustainable solutions in agricultural practices
DermAI: An Innovative AI-Driven Chatbot for Enhanced Dermatological Diagnosis and Patient Interaction
Get PDFSkin disorders constitute a noteworthy public health concern globally, with earnest impacts on both physical and mental well-being. However, effective dermatological care faces challenges in resource-limited regions due to poor infrastructure, limited access to medical facilities & expertise, and inadequate advanced diagnostic tools. The existing research work majorly focuses on cancer and uncommon skin diseases with models trying to achieve a higher training accuracy with no regards to misclassification rate. The products currently available in the market provide a limited initial diagnosis and suggest consulting a doctor to get an accurate diagnosis or offer a list of other possible skin disorders. To address these challenges, we propose DermAI, an innovative AI-based Chatbot made entirely of open-source technologies, which integrates the ResNet50 model and LLM via Chainlit, with Retrieval Augmented Generation(RAG), utilising AstraDB vector database and OpenAI embedding model for personalised responses. enabling accurate classification of common skin diseases. The proposed DermAI ensures minimal misclassification and comprehensive coverage of diseases, leveraging Retrieval-Augmented Generation and comparative model analysis. The metrics indicate that the model has a high true positive rate, with a misclassification rate of 2.17%, mean sensitivity, specificity & AUC of 92.6%, 99.8% & 99.9% respectively. This is demonstrated in the situations of melanoma, chickenpox, shingles, impetigo, and nail fungus, where it obtained 100% validation accuracy, a feat not attained by previous studies. Additionally, the model is highly capable of correctly identifying negative cases. The hallucination metric suggests the model may have a minimal tendency to hallucinate as the average hallucination score of 7% which falls far within the manually set threshold value of 50%. By setting the threshold value to 50%, the model generates grounded answers that are pertaining to the knowledge base and also allows it to be flexible with its responses. Overall, DermAI outperforms all solutions proposed in research literature