TELKOMNIKA (Telecommunication Computing Electronics and Control)
Not a member yet
3120 research outputs found
Sort by
Automatic diagnosis of rice plant diseases using VGG-16 and computer vision
Full text linkPathogens are organisms that cause disease in plants. In the case of rice, these pathogens can include fungi, bacteria, nematodes, protozoa, and viruses. This study aims to investigate rice plant diseases using a hybrid system that employs the visual geometry group-16 (VGG-16) architecture and computer vision techniques, alongside various optimization algorithms and hyperparameters. We utilize the convolutional neural network (CNN) architecture of VGG-16 for feature extraction, implementing a process known as transfer learning. Additionally, this research compares different optimization algorithms with the VGG-16 model to identify the most effective optimization for the CNN architecture applied to the tested dataset. The main contribution of this study is the development of a model for identifying rice plant diseases based on data collected using VGG-16 for feature extraction and neural networks for classification with specific parameters. Our findings indicate that the best optimization algorithm is stochastic gradient descent (SGD) with momentum, achieving training and validation loss results of 0.173 and 0.168, respectively. Furthermore, the training and validation accuracies were 0.95 and 0.957. The model’s performance metrics include an accuracy of 95.75, precision of 95.75, recall of 95.75, and an F1-score of 95.73
Enhancing spam detection using Harris Hawks optimization algorithm
Full text linkThis paper employs machine learning (ML) algorithms to identify and classify spam emails. The Harris Hawks optimization (HHO) algorithm can detect the crucial features that distinguish spam from ham emails. The HHO algorithm decreased the number of features in the ISCX-URL2016 spam dataset from 72 to 10. Implementing this will enhance the efficiency and cognitive acquisition of the ML algorithms. The decision tree (DT), Naive Bayes (NB), and AdaBoost algorithms are evaluated and contrasted to identify spam emails. The random search algorithm is used to optimize the significant hyperparameters of each algorithm for the specific task of spam identification. All three ML algorithms showed exceptional accuracy in detecting spam emails during the conducted testing. The DT algorithm attained a remarkable accuracy rate of 99.75%. The AdaBoost algorithm ranks second with an incredible accuracy of 99.67%. Finally, the NB algorithm attained an accuracy of 96.30%. The results demonstrate that the HHO algorithm shows promise in recognizing the crucial features of spam emails
Integrating artificial bee colony and cauchy algorithms for distribution network reconfiguration with soft open points
Full text linkReconfiguring the distribution network by selecting open switch states is an effective approach to reduce power losses in the system. However, with the rise of distributed energy resources such as photovoltaic and wind turbines and dynamic loads such as electric vehicles, which introduce uncertainties, it has become necessary to integrate standard operating procedures (SOPs) to better control power flows. This study proposes an algorithm that combines the artificial bee colony (ABC) and Cauchy opposition-based learning (OBL) algorithms to solve the optimization problem of determining both the location and capacity of SOPs, alongside reconfiguring the distribution network. The primary objective is to minimize power losses while improving power quality and system reliability. The proposed methodology was validated on the IEEE 33-node and 69-node distribution networks under seven varied operational scenarios, evaluating outcomes both with and without the integration of SOPs. The findings demonstrate that installing SOPs optimally reduces power losses, enhances system reliability, and maintains voltage levels within acceptable limits. The integration of the two algorithms also accelerates the convergence process, increasing computational speed and avoiding local optimization issues. When compared with other methods, the proposed algorithm delivers similar performance but with faster computation times and fewer iterations, making it more efficient and reliable
An architecture to build high performance infrastructures on cloud computing for telecommunications organizations
Full text linkNowadays, many small and medium organizations of the telecommunication sector must solve intrinsic heterogeneous problems in their own environments that have been associated with high computational complexities of their algorithms. These class of problems require to use high performance computing (HPC) infrastructures for their executions. Therefore, these must be accelerated to reduce significantly the execution times, included many problems that should be solved in real time: like the processing of multiples video streams, the pattern recognition in big volumes of data, the traffic analysis in cybersecurity solutions and among others. The building of HPC infrastructure permits to organize the technological platform to increase the productive and business indicators of the organizations. This paper describes an architecture as reference model and ecosystem for the building and systematic improvement of HPC infrastructures based on practical experiences from successive process of HPC infrastructure building on cloud deployment. That’s processes have been useful for the organizations permitting the integration of emergent hardware and software components launched to the international market. This landscape vision is pertinent for academics, scientifics and business organizations compelled to implement scientific and engineering applications to diverse fields that have a high impact in the society digital transformation
Lexicon-based comparison for suicide sentiment analysis on Twitter (X)
Full text linkSuicidal individuals frequently share their desires on social media. As a result, it was determined that a learning machine for early detection of suicide issues on social media was required. This study aims to examine Twitter (X) users’ suicide-related sentiment expressions. The results of searching X for the keywords ‘suicide’, ‘wish to die’, and ‘want to commit suicide’ for 4 months yielded 5,535 tweets. Following the cleaning process, 2,425 tweets were collected. The findings of labeling with the lexicon-based valence aware dictionary and sentiment reasoner (VADER) and Indonesia sentiment (INSET) lexicon, which psychologists confirmed, revealed that VADER was more accurate (92.1%) than INSET (81.6%). Sentiment research reveals negative (86.4%), positive (11.1%), and neutral (2.5%) sentiment. Support vector machine (SVM), K-nearest neighbor (KNN), and Naïve Bayes modeling results show accuracy above 86%, with SVM having the best accuracy (87.65%). Because of its great accuracy, this model can be used to identify and analyze suspicious behavior relating to suicide on X. Further research is still required, despite the excellent identification of early indicators of suicide ideation from social media posts
Advanced image processing techniques for intelligent building environments using pattern recognition
Full text linkThe use of smart building environments, along with high-technology image processing and pattern recognition, is discussed within this paper. The study shows that the Canny edge detection algorithm is better than the Sobel operator in the edge clarity, continuity and accuracy in segmenting those edges, posting 92.7% of edge detection accuracy. Incorporating fuzzy logic, the hybrid Hough transform, and sophisticated segmentation techniques, like adaptive simple linear iterative clustering (SLIC) superpixel division, the study advances line detection and feature identification in the images of buildings. The variational autoencoder (VAE) and principal component analysis (PCA) help optimise the feature extraction substantially by retaining more than 93% variance at a lower dimension. In addition, adaptive Otsu thresholding and region-growing segmentation allow improving the segmentation accuracy, resulting in a significant increase in building detection F1 score from 77.3% to 89.6%. Irrespective of the Hough transform issues like noise sensitivity and over-joining, the results suggest computing process ideas that are computationally effective, scalable, and applicable in smart building systems. This study suggests extending the current advancement of hybrid models and incorporating them with the urban planning procedures, energy control, and building security systems
Cost-effective long-range secure speech communication system for internet of things-enabled applications
Full text linkA new communication framework has been developed that allows voice transmission over long distances for internet of things (IoT) applications such as healthcare, smart cities, and remote monitoring in the least costly way and most secure manner. The system is based on long range (LoRa) technology and takes advantage of its spread spectrum technique, to provide long range transmission without the high-power requirements. The main limitation is LoRa’s bandwidth with a maximum throughput of 22 kbps for data. This presents a challenge for voice transmission communications. To address this shortened bandwidth issue, researchers developed an innovative compression solution that compresses voice data to less than 8 kbps to fit into LoRa’s capabilities. The compression allows for real practical voice communications and possibly can provide even greater distance than an uncompressed voice transmission update. The voice communications transmissions have cryptographic protection in place to protect the transmitted voice messages from unauthorized access
Improved classification for imbalanced data using ensemble clustering
Full text linkImbalanced datasets frequently occur in fields like fraud detection and medical diagnosis, where the number of instances in the majority class vastly exceeds those in the minority class. Traditional classification algorithms often become biased towards the majority class in these scenarios. To address this challenge, we introduce a novel method called improved classification using ensemble clustering (ICEC) for imbalanced datasets in this paper. ICEC merges classification with the strengths of consensus clustering to improve the classifier’s generalization ability. This approach utilizes a cluster ensemble to capture the structural characteristics of both the majority and minority classes, and the stable clustering scheme thus delivered is used to generate new auxiliary features. These features enhance the existing feature set, helping classifiers develop a more ro bust predictive model. Extensive testing on fifteen imbalanced datasets from the knowledge extraction based on evolutionary learning (KEEL) repository demonstrates the effectiveness of our proposed method. The approach was evaluated for random forest (RF) and linear support vector machine (SVM) classifiers on these data sets. Results indicate that ICEC proved to be effective for both classifiers, with an observed F1-score improvement of more than 10% for SVM and 3%for RF
Gain enhanced 5.8 GHz patch antenna with defected ground structure: design and measurement
Full text linkA rectangular microstrip patch antenna including a rectangular defective ground structure (DGS) is introduced to simultaneously enhance gain, bandwidth, and return loss while reducing antenna dimensions. This small antenna is engineered for 5.8 GHz applications, functioning throughout the frequency spectrum of 5.62 to 5.94 GHz. The design was executed on a 1.6 mm thick FR-4 substrate with a relative permittivity of 4.3, utilizing a microstrip line feed. The dimensions of the antenna are 31.75×28 ×1.6 mm³. The design approach utilized computer simulation technology (CST) Microwave Studio simulation software. The antenna attains resonance at 5.8 GHz, providing an initial bandwidth of 270 MHz and a return loss of -26 dB. A rectangular DGS was implemented to boost performance, yielding a 21.89% increase in bandwidth to 323 MHz and substantially enhancing the return loss from -23 dB to -47 dB. The gain increased from 3.95 dBi to 5.10 dBi, indicating a 30% enhancement, while sustaining an efficiency of around 83% at the resonant frequency. The antenna was constructed, and experimental measurements of parameters including gain and return loss closely matched the computer results
A convolution neural network model for knee osteoporosis classification using X-ray images
Full text linkBone structure deterioration along with low levels of bone density are the hallmarks of knee osteoporosis (KOP). The conventional approach for detecting osteoporosis is accomplished using a knee radiograph, but it requires specialized knowledge. Nevertheless, X-rays can be difficult to interpret due to their large volume and minor fluctuations. In the past few decades, deep learning algorithms have minimized misinterpretation and modified medical diagnosis. In particular, algorithms based on convolutional neural networks (CNNs) have been used to speed up the procedure of diagnosis because of their innate capacity to extract significant features that often are challenging to spot by hand. A robust CNN model was proposed in this paper for KOP classification which uses a train and test approach to recognize healthy, osteopenia-predicted, and osteoporosis knee cases using 1947 X-ray images. The proposed model was designed using Jupyter Notebook and is in Python. To verify the efficiency of the model, some factors were calculated such as accuracy, precision, recall, and f1-score. In comparison with other similar systems, the results obtained showed that the accuracy of the proposed system reached 90.25%