Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
Forecasting bitcoin price fluctuations: a time series analysis approach for predictive modelling
Get PDFThe recent fluctuations in the price of Bitcoin, marked by both significant increases and subsequent decreases, has attracted media and public attention. Consequently, many researchers have explored various factors influencing Bitcoin’s price and the underlying patterns of its fluctuations. This paper aims to predict and analyses the factors affecting Bitcoin’s price by creating a unique dataset with nearly 40 features and deriving two child datasets using correlation and mutual information as feature selection techniques. Additionally, we train machine learning models, including linear regression (LR), extreme gradient boosting (XGBoost), support vector regression (SVR), Facebook Prophet (FB Prophet), and bidirectional gated recurrent unit (BI-GRU), to predict Bitcoin’s next-day price. The model’s performance is evaluated using mean square error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and R2 score metrics. Our findings indicate that machine learning techniques are effective in predicting Bitcoin’s price and could be valuable for investors seeking to maximize profits
Advancing supply chain management through artificial intelligence: a systematic literature review
Get PDFThis study evaluates the role and impact of artificial intelligence (AI) in supply chain management (SCM). Following a five-step process, the review covered academic publications from 2000 to 2024, drawing from different databases. The review identified 426 relevant articles for analysis, focusing on AI techniques. The analysis explored their applications, advantages, and barriers to adoption in SCM. The study also discussed key challenges, including financial, organizational, strategic, technological, and legal barriers. The findings suggest that while AI techniques offer significant potential for improving SCM, several obstacles hinder their broader implementation. Addressing these obstacles requires investments in infrastructure, skills development, and effective change management
RNN Diabetic framework for identifying diabetic eye diseases
Get PDFMany areas of image identification and classification for medical imaging diagnostics have greatly benefited from deep learning (DL). Diabetic retinopathy (DR) will become the most common cause of blindness worldwide, making diabetes a major threat to public health. This research proposes an automated identification system using deep recurrent neural networks (RNNs) to identify and classify four categories of diabetic eye diseases: DR, cataract, glaucoma, and diabetic macular edema (DME). We use three different model architectures based on RNN and their types, we called our proposed system RNN Diabetic framework. These models are combined with one of the commonly used architectures that support sufficient accuracy and speed for the model which is residual network (ResNet)152V2. The three model architectures are RNN+ResNet152V2, gated recurrent unit (GRU)+ResNet152V2, and bidirectional GRU (Bi-GRU)+ResNet152V2. The proposed models were assessed as collected datasets: DIARETDB0, DIARETDB1, messidor, HEI-MED, ocular, and retina. A full analysis and evaluation of these three deep RNN architectures are presented. The experiments showed that the Bi-GRU+ResNet152V2 model worked better than the other two proposed models. In addition, we compare these three proposed models with the previous studies and find that the proposed Bi-GRU+ResNet152V2 model achieves the highest results with accuracy equal to 99.8%, 98.1% sensitivity, 98.6% specificity, 99.8% precision, 99.8% F1 score, and 99.8% areas under the curve (AUC)
A hybrid firefly algorithm for the sales representative planning problem
Get PDFIn the rapidly increasing pharmaceutical sector, sales representatives are employed by pharmaceutical manufacturers and distributors to inform and educate physicians. To convince providers to prescribe the medications to their patients, these representatives rely on their product expertise and people’s abilities to close deals. Instead of making direct sales, pharmaceutical sales representatives help medical professionals get the medications, treatments, and information they need to give their patients the best care possible. Furthermore, they inform the public about novel and occasionally life-saving treatments and share interesting medical developments. This study presents a hybrid methodology that integrates the benefits of local search and the firefly algorithm (FA) to determine the optimal planning for a sales representative. The objective is to maximize the rewards while adhering to certain constraints. The objective is to maximize the rewards while adhering to certain limits. Utilizing local search, the hybrid algorithm enhances firefly’s global search behaviour and produces the best possible sales presentation planning. The experimental findings demonstrate the superior performance of the suggested algorithm compared to the FA and other literature methods in the sense of enhancing the convergence rate and preventing local minima. Furthermore, it can enhance the best-known solution for most benchmark instances
A novel RGB image steganography algorithm using type-1 fuzzy logic
Get PDFSteganography aims to conceal secret data within images without affecting image quality. Traditional methods often struggle with balancing simplicity, effectiveness and payload capacity while maintaining imperceptibility. Proposed algorithm: the paper proposed a novel steganographic mshEdgeRGB_T1 algorithm that combines Mamdani fuzzy type-1 logic with the least significant bit (LSB) method. The LSB method is chosen for its simplicity and effectiveness in hiding messages. The mshEdgeRGB_T1 algorithm focuses on embedding secret messages in edge pixels, detecting more edge pixels compared to other methods, thus increasing payload capacity. Findings: the algorithm’s performance is evaluated using metrics such as peak signal-to-noise ratio (PSNR), mean squared error (MSE) and histogram analysis to measure the similarity between the cover and Stego images, quantifying the level of imperceptibility. Experimental analysis demonstrates that the mshEdgeRGB_T1 algorithm offers improved payload capacity, enhanced security and reduced imperceptibility compared to many existing methods. Conclusion: the proposed mshEdgeRGB_T1 algorithm effectively balances simplicity, payload capacity and image quality, making it a better use for image steganography
Android malware detection through opcode sequences using deep learning LSTM and GRU networks
Get PDFAndroid malware detection was a complex task due to the intricate structure of Android applications, which consisted of numerous Java methods and classes. Effective detection required the extraction of meaningful features and the application of advanced machine learning (ML) or deep learning (DL) algorithms. This paper presented a novel approach to detecting Android malware by leveraging opcode sequences extracted from Android applications. These opcode sequences, which differed between malicious and benign apps, formed the basis of the detection model. The methodology involved extracting opcode sequences from decompiled Android APK files using the “Androguard” tool and applying recurrent neural networks (RNN) with long short-term memory (LSTM), Bi-LSTM, and gated recurrent unit (GRU) architectures to classify the apps as either malware or benign. The combination of these advanced DL techniques allowed for capturing temporal dependencies in opcode sequences, resulting in a significant improvement in detection capabilities. This work underscored the potential of using opcode sequences in conjunction with RNN, LSTM, and GRU for robust and accurate malware detection, while also highlighting the importance of further exploring additional features for comprehensive classification
Virtual learning environment on satisfaction and academic performance of students in institutions of higher learning
Get PDFAs a result of the COVID-19 outbreak in 2020, education institutions across the world had to come to a functional standstill since they had to protect their students from viral exposures thereby affecting academic activities. However, several institutions had to adopt online virtual learning environments (VLE) using basic information and communication technology tools to provide platforms for teaching and learning thereby mitigating the effects of the pandemic on the students. This study was focused on the identification of the various types of VLE tools that were adopted alongside the impact that these tools had on learning satisfaction and the academic performance of students of higher learning in Nigeria. This study adopted a purposive simple random selection of undergraduate students of the department of computer science who had adopted the use of VLE to learn during the period of the pandemic. The results of the study showed that the most popular VLE tools were Zoom, Google Classroom, WhatsApp, Telegram, Coursera, Google Forms and learning management systems (LMS) while the least popular VLE tools were Microsoft Teams, Moodle/Edmondo, and Google Meet. The results showed that the students agreed to their behavioral intention to use VLE, the impact of VLE on learning satisfaction, and the impact of VLE on academic performance alongside the existence of a positive correlation among the research variables
Data mining and cardiac health: predicting heart attack risks
Get PDFIn a context where heart attacks continue to be a global health concern, the lack of precision in predicting who is at higher risk poses a critical challenge due to the variability of risk factors and complex interactions among them. The research aims to develop predictive models for heart attack risks using data mining techniques, employing the knowledge discovery in databases methodology (KDD) and the k-means algorithm with RapidMiner studio. The primary objective is to identify patterns and risk profiles, allowing for early identification of at-risk individuals, considering factors like obesity, diabetes, alcoholism, and stress, to reduce preventable deaths and improve cardiac healthcare. This innovative approach combines cardiac health, data mining, and KDD methodology to address the challenge of predicting heart attack risks and has the potential to enhance medical care and save lives. The predominant results obtained were that cluster 1 with a fraction of 0.312 and a percentage of 31.2% of the attribute diabetes was one of the most prevalent causes of cardiac risk. Finally, the research concluded that people with diabetes are more likely to have cardiac risk associated with dietary factors or consumption of other substances
TextBugger: an extended adversarial text attack on NLP-based text classification model
Get PDFRecently, adversarial input highly negotiates the security concerns in deep learning (DL) techniques. The main motive to enhance the natural language processing (NLP) models is to learn attacks and secure against adversarial text. Presently, the antagonistic attack techniques face some issues like high error and traditional prevention approaches accurately secure data against harmful attacks. Hence, some attacks unable to increase more flaws of NLP models thereby introducing enhanced antagonistic mechanisms. The proposed article introduced an extended text adversarial generation method, TextBugger. Initially, preprocessing steps such as stop word (SR) removal, and tokenization are performed to remove noises from the text data. Then, various NLP models like Bi-directional encoder representations from transformers (BERT), robustly optimized BERT (ROBERTa), and extreme learning machine neural network (XLNet) models are analyzed for outputting hostile texts. The simulation process is carried out in the Python platform and a publicly available text classification attack database is utilized for the training process. Various assessing measures like success rate, time consumption, positive predictive value (PPV), Kappa coefficient (KC), and F-measure are analyzed with different TextBugger models. The overall success rate achieved by BERT, ROBERTa, and XLNet is about 98.6%, 99.7%, and 96.8% respectively
Design and development of an automated spirulina (Arthrospira platensis) algae cultivator
Get PDFThe cultivation of algae has gotten more attention from algae enthusiasts who have seen the benefits of algae in many uses. To maximize productivity, the parameters for growth of this algae must be controlled, such as pH, turbidity, light intensity, and the mixture solution for optimal growth. In this paper, an automated spirulina algae cultivator is designed and developed in a small-scale pond to replace the existing manual process. The system developed is composed of compact and modular cultivation unit, ph sensor, water level sensor, turbidity sensor, light intensity sensor, and motor actuators for mixing solutions. Each parameter was controlled individually in an on-off control system. A simple nutrient addition program (SNAP) solution is also used for better growth productivity by maximizing its nutrient contents. The pH is maintained at 9 to 12 for a healthy biomass output. Daily weight measurement was conducted using an analytical balance to monitor its growth. Using the developed prototype recorded a 33% higher rate of productivity over the manual process. This setup can potentially be used as a model for mass production of spirulina algae