Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
A novel CNN-ANN fusion approach for improved facial emotion detection
Get PDFIn recent years, the field of emotion recognition has witnessed an increased interest due to the rise of deep learning techniques. However, one of the persistent difficulties in this domain, which we have attempted to address, is the variability in image sizes utilized. In this study, we have reviewed the work by different researchers and summarized their key findings. In our research, we introduce a novel technique that integrates the strengths of 1D convolutional neural networks (CNNs) and artificial neural networks (ANNs) through a late fusion model, leveraging CNNs' shared weights and automatic feature learning for spatial and temporal data, alongside ANN's comprehensive feature consideration. Our research findings highlight the effectiveness of this approach, which achieves a remarkable accuracy of 92.42%, along with other evaluation metrics demonstrating notable results. Furthermore, we conduct a comprehensive analysis of the proposed method, comparing it with advanced methods in the field of facial emotion recognition. Through this comparative analysis, we demonstrate the superiority of our proposed approach, addressing challenges that have not yet been addressed till date, thus leading to progress in this field
Advancing chronic pain relief cloud-based remote management with machine learning in healthcare
Get PDFHealthcare providers face a significant challenge in the treatment of chronic pain, requiring creative responses to enhance patient outcomes and streamline healthcare delivery. It suggests using cloud-based remote management with machine learning (ML) to alleviate chronic pain. Wearable device data, electronic health record (EHR) data, and patient-reported outcomes are all inputs into the suggested system’s data analysis pipeline, which combines support vector machines (SVM) with recurrent neural networks (RNN). SVM’s powerful classification skills make it possible to classify patients’ risks and predict how they will react to therapy. RNNs are very good at processing sequential data, which means they may identify trends in patient symptoms and drug adherence over time. By integrating these algorithms, healthcare professionals may create individualized treatment programs that consider each patient’s preferences and specific requirements. Early intervention and proactive treatment of pain symptoms are made possible by the system’s ability to monitor patients in real-time remotely. The system is further improved by using predictive analytics to identify patients who could benefit from extra support services and to forecast when they will have acute pain episodes. The proposed approach can change the game regarding managing chronic pain. It provides data-driven, individualized treatment that improves patient outcomes while cutting healthcare expenses
Sentiment analysis based on Indonesian language lexicon and IndoBERT on user reviews PLN mobile application
Get PDFPLN mobile application as an integrated platform for self-service among mobile consumers, facilitating easier access to various services, including receiving information such as public complaints. The application can be downloaded through the Google Play Store and App Store, and users can express their opinions through reviews and ratings. In this era of advanced technology, aspects such as reviews, ratings, and evaluations have important value for business practitioners. However, there are often inconsistencies between ratings and reviews that do not fully represent the quality of the application. In response, a study was conducted to analyze the sentiment of user reviews from January to June 2022, by collecting 1,000 review samples from the Google Play Store. The data was collected using web scraping techniques and then processed into a dataset through text pre-processing methods. Sentiments were analyzed using an automatic labeling method in Indonesian based on a lexicon known as INSET (Indonesia sentiment), which resulted in 482 positive reviews, 144 negative reviews, and 374 neutral reviews. The next step is classification using Indonesian bidirectional encoder representations from transformers (IndoBERT). In this process, the data was divided into testing, training, and validation sets with a ratio of 80:10:10. The analysis managed to achieve an impressive accuracy rate of 81%
Integration of message queue and drop policy in spray and hop distance protocol for DTNs in smart city scenario
Get PDFImplementing delay tolerant networks (DTNs) in smart cities for developing countries is promising. DTNs offer a low-cost network communication solution without expensive infrastructure, such as 3G/4G or LPWA networks provided by commercial operators. This paper considers sensor data collection in the Surabaya smart city scenario, and the use of DTNs is examined instead of the expensive infrastructure. However, a customized and optimized protocol is required since no existing DTN routing protocol perfectly matches the scenario. This paper proposes integrating a hop count-based message queue (HCMQ) and a message time-to-live (TTL)-based drop policy (MTDP) into the spray and hop distance protocol (SNHD), an enhanced version of the spray and wait (SNW) protocol. The Surabaya smart city scenario was simulated on the one simulator with a wide range of message generation rates at each sensor node. The proposed integration significantly improves the total size of delivered messages, especially when the message generation rate is high, i.e., in congested situations, compared to other routing protocols in this scenario. It also exhibits an average latency lower than other routing protocols. Overall, this integration enhances the DTNs protocol’s performance in a low-cost alternative data collection in the Surabaya smart city scenario
Classification of traditional Joha rice using deep neural network
Get PDFRice is the major food for the most of the people in the world. It is grown comfortably in a rain fed area. Rice is called a kharif crop in India. According to the International Rice Research Institute (IRRI) classification rice is classified as Ahu, Sali, Boro and Hill rice. In north eastern India the Joha rice is mainly cultivated in the state of Assam. There are many varieties of Joha rice like Kunkuni, Rampal, Manipuri, Tulshi, and Keteki. But it is quite complex for the farmers as well as the common people to differentiate these type of Joha rice due to their morphological structure. Initially this type of classification done by naked eyes or used some laboratory experiments. But due to the tiredness and some external factors it is not fruitful. Some of the image classification technologies are used in classification as well identification of rice in different research work using MATLAB. At present deep learning plays a tremendous role in image analysis in agricultural domain. Here in this paper we take two verities of Joha rice which are closer in morphological structure and cannot be separated by our naked eyes. Using deep neural network, binary classification is done in these Joha rice
An implementation of GAN analysis for criminal face identification system
Get PDFIn recent times, the criminal activities are growing at an exponential rate. For the prevention of crime, one of the main issues that are before the police are accurate identification of criminals and on the other hand the availability of police officers are not adequate. The most tedious task is tracking the suspect once a crime was committed. Over the years, several technical solutions have been presented to detect the criminals however most of them were not effective. One of the most significant characteristics for the identification of a person is face. Even identical twins have their own unique faces. Face identification is a challenging topic in computer vision because the human face is a dynamic entity with a high degree of visual variation. In this area, identification accuracy and speed are significant challenges. Hence to solve these issues, an implementation of generative adversarial network (GAN) analysis for criminal face identification system is presented. GAN is used for the identification of criminals. Recall, precision, accuracy, and F1-score are used to assess the performance of the presented technique. Compared to previous models, this model will achieve better performance for criminal face detection
Mechanized network based cyber-attack detection and classification using DNN-generative adversarial model
Get PDFThese days almost everything is internet. Cyberattacks are the world's most pressing issues. Due to these attacks, Computer systems can be rendered inoperable, disrupted, destroyed or controlled via cyberattacks. Additionally, they can be used to steal, modify, erase, block, or alter data. Most organizations are facing this Issue and lose financially as well as in data security, there are numerous conventional intrusion detection systems (IDS) and firewalls are illustrations for network security tools which are not able to classify and detect different types of attacks in network. With machine learning approach using the Dataset KDD_CUP 99 as input, the synthetic minority oversampling technique (SMOTE) is one of the most often used oversampling methods for addressing imbalance issues. The proposed hybrid deep neural network (DNN), generative adversarial network (GAN), and exhaustive feature selection (EFS) can detect and classify several attack types including R2L, U2R, Probe, denial of service (DoS), and normal attacks types and inform to administrator to ring alarm sound to control and monitor network traffic in dynamically typed networks
Crop prediction using an enhanced stacked ensemble machine learning model
Get PDFIn India, agriculture is a major sector that fulfils the population's food requirements and significantly contributes to the gross domestic product (GDP). The careful selection of crops is fundamental to maximizing agricultural yield, thereby elevating the economic vitality of the farming community. Precision agriculture (PA) leverages weather and soil data to inform crop selection strategies. Conventional machine learning (ML) models such as decision trees (DT), support vector classifier, K-nearest neighbors (KNN), and extreme gradient boost (XGBoost) have been deployed to predict the best crop. However, these model's efficiency is suboptimal in the current circumstances. The enhanced stacked ensemble ML model is a sophisticated meta-model that addresses these limitations. It harnesses the predictive power of individual ML models, stratified in a layered architecture to improve the prediction accuracy. This advanced model has demonstrated a commendable accuracy rate of 93.1% prediction by taking input of 12 soil parameters such as Nitrogen, Phosphorus, Potassium, and weather parameters such as temperature and rainfall, substantially outperforming the accuracies achieved by the individual contributing models. The efficacy of the proposed meta-model in crop selection based on agronomic parameters signifies a substantial advancement, fortifying the economic resilience of India's agriculture
Discrete wavelet transform and convolutional neural network based handwritten Sanskrit character recognition
Get PDFSanskrit is one of the ancient languages from which the majority of present Indian languages are developed. Although the national mission for manuscripts (NMM) is digitizing handwritten Sanskrit manuscripts, there are still a lot of papers that need to be digitized. Recognition of handwritten script is a challenging task due to individual differences in writing styles and how those variations alter over time. The Sanskrit language is written in Devanagari script. A novel approach using discrete wavelet transform (DWT) and convolutional natural network (CNN) is proposed in this paper. Devanagari handwritten character dataset which includes 2000 handwritten images of 36 classes (2000*36=72000) is used in this research. Fine-tuned GoogLeNet model implemented here gave optimum values of epochs and learning rate of 15 and 0.01 respectively. Classification accuracy obtained by proposed DWT – CNN model is 98.97% with a loss of 0.098. Fine-tuned GoogLeNet model achieves 99.68% accuracy with a 0.0635 loss. Results obtained are also compared with existing approaches and found superior
Predict glucose values with DE algorithm optimized T-LSTM
Get PDFThe prevalence of diabetes is rising. According to the International Diabetes Federation (IDF) predictions, the number of diabetic patients worldwide will reach 608 million in 2030, accounting for approximately 11.3% of the total number of people in the world. To monitor and predict the future 1 hour glucose have a great significance meaning for patients. This research utilizes a differential evolution (DE) algorithm, an optimized hybrid model transformer and long short-term memory (T-LSTM) technologies to analyze historical data from continuous blood glucose monitoring (CGM) systems and equipment calibration values. The aim is to predict future blood sugar levels in patients, thereby helping to prevent episodes of hypoglycemia and hyperglycemia. The study tested the model using the CGM data from 8 patients at the Suzhou Municipal Hospital in Jiangsu Province, China. Results show that this DE-optimized T-LSTM model outperforms traditional models. The model's accuracy is evaluated using mean squared error (MSE), with MSE values recorded at 15, 30, and 45 minutes being 0.96, 1.54, and 2.31, respectively