International Journal of Informatics and Communication Technology (IJ-ICT)
Not a member yet
494 research outputs found
Sort by
Advanced predictive models for thyroid disease comorbidities using machine learning and deep learning: a comprehensive review
Get PDFWith advances in machine learning (ML) and deep learning (DL), the future of thyroid disease diagnosis and prognosis looks very bright. The integration of various data such as imaging and medical record data has increased the accuracy of the model. Advanced DL models such as convolutional neural network (CNN) and recurrent neural network (RNN) further improved disease detection in precision medicine. However, some of the major disadvantages of effective clinical integration include unbalanced samples, unclear sampling, having to communicate in different populations, decreased physician confidence due to the vagueness of current models therefore, and few studies available to identify thyroid comorbidities such as polycystic ovary syndrome (PCOS) and thyroid eye disease (TED) in a variety of different populations to develop the line. It is important to focus future research activities on model definition and validation an improving and thus the diagnosis and prognosis of thyroid comorbidities is of utmost importance. What this will bring is ML and DL, an opportunity to make very significant improvements in the diagnosis, treatment, and management of thyroid diseases, thereby improving patient outcomes and health care by seeking crystals as a group they work interdisciplinary to collaborate in developing flexible solutions, sharing knowledge, and responding to these stated deficiencies
Automatic identification of native trees using MobileNetV2 model
Get PDFIn protecting our biodiversity, knowledge of tree species is vital. However, not all people are familiar with the trees present in the community which can affect their ability to fully protect the trees. In this premise that the researchers decided to conduct this study to support the sustainable forest management project in the Province of Quirino through the creation of a model of automatic identification of native trees, using the leaves of the trees, found within the Quirino Forest landscape. The model aims to help residents with accessible tools for tree identification which can be used in the conservation efforts within the province. Transfer learning for deep learning, one of the latest advancements in image processing, shows potential for tree identification because the method dodges the labor intensive feature engineering. Using the Quirino Province native trees leaf/leaflet images dataset, which was annotated by foresters, the MobileNetV2 convolutional neural network was evaluated systemically in this paper. The result shows that the best model version to classify the native trees based on their leaves or leaflets is the one produced using 800 training steps which yields an overall accuracy of 89.61%. The result attained for the tree identification indicates that the proposed technique might be an appropriate tool to assist humans in the identification of native trees found within the landscape of Quirino and can provide reliable technical support for sustainable forest management
Revolutionizing human activity recognition with prophet algorithm and deep learning
Get PDFVarious industries, such as healthcare and surveillance, depend heavily on the ability to recognize human activity. The “human activity recognition (HAR) using smartphones data set” can be found in the UCI online repository and includes accelerometer and gyroscope readings recorded during a variety of human activities. The accelerometer and gyroscope signals are also subjected to a band-pass filter to eliminate unwanted frequencies and background noise. This method effectively decreases the dimensionality of the feature space while improving the model's accuracy and efficiency. “Convolutional neural networks (CNNs)” and “long shortterm memory (LSTM)” networks are combined to create pyramidal dilated convolutional memory network (PDCMN), which is the final proposal. Results from experiments demonstrate the effectiveness and reliability of the suggested method, demonstrating its potential for precise and effective HAR in actuality schemes
Novel multilevel local binary texture descriptor for oral cancer detection
Get PDFCategorizing texture medical images is an extensive job in most of the fields of computer vision, pattern recognition and biomedical imaging. For the past few years, the texture analysis system model, especially for biological images, has been brought to attention because of its ever-growing requirements and characteristics. This research shows its novelty by using a multilevel local binary texture descriptor (MLBTD) algorithm with support vector machine (SVM), k-nearest neighbor (KNN), and CT Classifiers to investigate the texture features of the oral cancer samples. The simulation work is done in MATLAB2021a environment by employing the MLBTD algorithm. A Mendeley dataset, containing 89 oral cavity histopathological images and 439 OSCC images in 100x magnification, is used. A statistical comparative study of local binary pattern (LBP) and MLBTD with linear SVM, KNN, CT classifier is performed in which results show the better performance of MLBTD and linear SVM with 89.94% of accuracy and by applying MLBTD algorithm over 90.57% accuracy is obtained whereas LBP algorithm only provides 86.16% of accuracy
Digital control of plant development through sensors and microcontrollers in Kosova
Get PDFThe plant monitoring system aims to develop an automated solution for optimizing plant growth. Using the Arduino Uno ATMEGA328P microcontroller module and various sensors, this system regulates environmental conditions to promote optimal plant development. It requires adequate software to operate effectively, enabling the microcontroller to monitor and regulate climatic conditions. The primary goal of this paper is to present a comprehensive system that continuously measures parameters such as light intensity, air humidity, and soil moisture in real time within a vegetable greenhouse or a plastic-covered plant environment. This scientific paper provides an in-depth description of the hardware components used, their electronic connections, and the implementation of program code written in C++. Based on the measured physical parameters, the plant monitoring system performs specific actions, such as watering the plants and regulating the ambient temperature. In conclusion, this system effectively supports healthy plant growth and enhances the quality and yield of plant products. The paper serves as a practical example for improving plant cultivation in the agricultural sector in the Republic of Kosova
A hybrid approach using VGG16-EffcientNetV2B3-FCNets for accurate indoor vs outdoor and animated vs natural image classification
Get PDFThe paper introduces a hybrid approach that synergistically combines the strengths of VGG16, EfficientNetV2B3, and fully connected networks (FCNets) to achieve precise image classification. Specifically, our focus lies in discerning between basic indoor and outdoor scenes, further extended to distinguish between animated and natural images. Our proposed hybrid architecture harnesses the unique characteristics of each component to significantly enhance the model’s overall performance in fine-grained image categorization. In our methodology, we utilize VGG16 and EfficientNetV2B3 as the feature extractors. During evaluation, we examined various classification algorithms, such as VGG16, EfficientNet, Feature_Aggr_Avg, and Feature_Aggr_max, among others. Notably, our hybrid feature aggregation approach demonstrates a remarkable improvement of 0.5% in accuracy compared to existing solutions employing VGG16 and EfficientNet as feature extractors. Notably, for indoor versus outdoor image classification, feature_aggr_avgachieves an accuracy of 98.51%. Similarly, when distinguishing between animated and natural images, Feature_Aggr_Avgachieves an impressive accuracy of 99.20%. Our findings demonstrate improved accuracy with the hybrid model, proving its adaptability across diverse classification tasks. The model is promising for applications like automated surveillance, content filtering, and intelligent visual recognition, with its robustness and precision making it ideal for realworld scenarios requiring nuanced categorization
Performance analysis of D2D network in heterogeneous multitier interference scenarios
Get PDFThe trade-off between boosting network throughput and minimizing interference is a critical issue in fifth generation (5G) networks. Diverting the data traffic around the network access point in device-to-device (D2D) communication is an important step in realizing the vision of 5G. Though the D2D network improves the network performance, they complicate the interference management process. Interference is an invisible physical phenomenon occurring in wireless communication which happens when multiple transmissions happen simultaneously over a general wireless medium. Enormous growth in usage of mobile phone and other wireless gadgets in recent years has paved the way for significant role in Interference analysis over multi-tier network. Interference could affect communication systems performance and it might even prevent systems functioning properly. 3G and 4G wireless devices coexisted with reverse compatibility in a coverage area. Also, after their widespread adoption, 5G devices have become prevalent across the globe and this reaffirms interference coexistence as a significant field of research. Multiple systems operating in a region will cause severe interference and ultimately reduce the quality of received signal. A simulation environment for cellular standards coexistence considering real-time parameters is created and experimented. Various research works earlier addresses the interference mitigation techniques in multi-tier networks but none of them present the analysis of scenarios and interfering signal power levels in the respective contexts. In this paper various scenarios with different network interference coexistence were studied, simulated, and analyzed quantitatively
Modeling chemical kinetics of geopolymers using physics informed neural network
Get PDFUsing a physics informed neural network for the analysis of geopolymers as an alternate material for cement can be a viable approach, as neural networks are capable of modeling complex, nonlinear relationships in data, which can be beneficial for representing the dynamics of chemical properties. If you have a substantial amount of theoretical data, a neural network can learn patterns and relationships in the data, even when the underlying system dynamics are not well-defined or are difficult to model analytically. A welltrained neural network can generalize from the training data to make predictions for unseen scenarios, which can be useful for real-time analysis of the material
Machine learning in detecting and interpreting business incubator success data and datasets
Get PDFThis research contributes to creating a proposed architectural model by utilizing several machine learning (ML) algorithms, heatmap correlation, and ML interpretation. Several algorithms are used, such as K-nearest neighbors (KNN) to the adaptive boosting (AdaBoost) algorithm, and heatmap correlation is used to see the relationship between variables. Finally, select K-best is used in the results, showing that several proposed model ML algorithms such as AdaBoost, CatBoost, and XGBoost have accuracy, precision, and recall of 94% and an F1-score of 93%. However, the computing time the best ML is AdaBoost with 0.081s. Then, finally, the proposed model results of the interpretation of AdaBoost using select K-best are the best features “last revenue” and “first revenue” with k feature values of 0.58 and 0.196, these features influence the success of the business. The results show that the proposed model successfully utilized model classification, correlation, and interpretation. The proposed model still has weaknesses, such as the ML model being outdated and not having too many interpretation features. The future research might maximize with ML models and the latest interpretations. These improvements could be in the form of ML algorithms that are more immune to data uncertainty, and interpretation of results with wider data
Deep learning algorithms for breast cancer detection from ultrasound scans
Get PDFBreast cancer is a highly dangerous disease and the leading cause of cancer related deaths among women. Early detection of breast cancer is considered quite challenging but can offer significant benefits, as various treatment interventions can be initiated earlier. The focus of this research is to develop a model to detect breast cancer based on ultrasound results using deep learning algorithms. In the initial stages, several preprocessing processes, including image transformation and image augmentation were performed. Two types of models were developed: utilizing mask files and without using mask files. Two types of models were developed using four deep learning algorithms: residual network (ResNet)-50, VGG16, vision transformer (ViT), and data-efficient image transformer (DeiT). Various algorithms, such as optimization algorithms, loss functions, and hyperparameter tuning algorithms, were employed during the model training process. Accuracy used as the performance metric to measure the model’s effectiveness. The model developed with ResNet-50 became the best model, achieving an accuracy of 94% for the model using mask files. In comparison, the model developed with ResNet-50 and DeiT became the best model for the model without mask files, with an accuracy of 80%. Therefore, it can be concluded that using mask files is crucial for producing the best-performing model