Earticle

Recent solutions of storing and distribution of patient data have a number of confines that limits users access to their patient health records(PHR), decrease accessibility of critical information to care providers, and eventually extant a boundary in the transfer of traditional healthcare into a digital healthcare approach. In order to remediate such shortcomings blockchain based solution is the best tool to provide storage and trust. In digital healthcare data allocation, numerous cloud based approaches have been suggested, but it would be untrustworthy to rely on the third-party. In recent times, blockchain has been implemented in digital healthcare record sharing, which skip to trust on a third party. Although, current methods only focus over the clinical related records received from medical diagnosis. They are not considered resourceful regarding its data sharing which are always generated from biomedical and monitoring devices. In this research we have proposed blockchain as a novel approach to secure patient related data access, implementation obstacles, and a strategy for transitioning gradually from current technology to a blockchain based solution. Keywords: Blockchain, healthcare, patient health record, hyperledger fabric, data sharing, security, decentralization, trust chain.

The main research topic of this study is how much ‘opinion mining’ of online comments on specific keywords reflects actual public opinion. In detail, we compared and analyzed how much the results of sentiment analysis for comments by platform reflect the actual opinion poll results. We analyzed the most mentioned keywords by platform and by parking in the comments classified as positive, and the most mentioned keywords by platform and by parking in the comments classified as negative. As a result of the study, it was found that the results of the polls were similarly reflected in the order of the Naver News model, Naver News + YouTube model, and YouTube model. In addition, it was possible to find out keywords with high interest by positive/negative public opinion through positive/negative word cloud analysis by parking and platform.

Sewer pipes are an essential public infrastructure of countries worldwide. They support wastewater transportation for processing or disposal. The harsh environments inside the sewer pipes can lead to the occurrence of various defects. Current crack detection approaches mainly focus on the surveillance camera (CCTV) to assess the condition of the sewer pipes. This process is considered a tiresome and laborious process. Therefore, a robust and efficient sewer defect detection system based on the transformer architecture is introduced in this manuscript. In addition, the system can provide explainable visualization for its predictions using the transformer's attention.

With the recent development of computer technology, machine learning is being applied to the semiconductor field. However, it takes a lot of time in existing simulators to make a large number of samples required for this. In this paper, samples were produced in two ways to obtain samples in a short time. The similarity between the data of the generated samples and the real data distribution was measured and compared.

Digital assets are one of the most important precious entities for any organization and if someone captures them for the purpose of ransom, then it would be a serious threat. The threat actor behind this activity is the ransomware. The threat posed by the ransomware on personal and business data assets expands very quickly. Data on an infected computer becomes encrypted until a ransom is paid for its release. Each year, ransomware causes hundreds of millions of dollars of losses for the companies throughout the world. Frequently, new versions are released because of the enormous profit margins and notorious practices. Antivirus software and other intrusion detection systems can be bypassed, so they are not a permanent solution so-far. This research work contributes some latest dissemination and mitigation techniques that are using in ransomware attacks. We also discussed the countermeasures to mitigate the ransomware attacks and some decryption tools and ransomware simulation to find the vulnerabilities in the system.

Breadth-first search (BFS) is one of the most used graph kernels, and substantially affects the overall performance when processing various graphs. Since graph data are frequently used in real life for example road networks in navigation systems, high performance graph processing becomes more critical. In this study, we aim to process BFS algorithm efficiently on road network data. We propose BARON, a BFS algorithm that copes with road networks. To accelerate graph traversal, BARON reduce the occurrence of branch and memory divergences by exploiting warp-cooperative work sharing and atomic operations. With this design approach, BARON outperforms the other BFS kernels of state-of-the-art graph processing frameworks executed stably on the latest GPU architectures. For various graphs, BARON yields speedups of up to 2.88 and 5.43 over Gunrock and CuSha, respectively.

Fire is an extremely catastrophic disaster that leads to the destruction of forests, human assets, reduced soil fertility, land resources, and the cause of global warming. In the current decade, fire detection and its management are the major concern of several researchers to prevent social, ecological, and economic damages. To overcome such kind of losses, early fire detection, and the automatic response is very significant. Moreover, achieving high accuracy with reducing inference time and model size is also challenging for the Unmanned Aerial Vehicle (UAVs). Therefore, in this work, we enabled the VGG16 architecture for UAV in terms of reducing its learning parameters from 138 million to 11.4 million for early fire detection. The proposed system is inexpensive in terms of computation and size. The performance of our proposed work is evaluated over the custom dataset. We performed comprehensive experiments using various deep learning architectures such as VGG16, ResNet50, and the proposed CNN model. The experimental results based on the proposed model achieved an accuracy of 98% on 50 epochs.

We address data collection issues in monocular 3D object detection. Since deep learning based object detection systems often require a variety of human annotations for advanced functionalities and also a lot of training datasets for better performance, it is increasingly important for researchers to review currently available databases, and prepare their own datasets with customized labels for target applications such as monocular 3D object detection. On top of the survey of related datasets, we study two different methods of collecting new datasets. As an example of our methods, we present five realworld object instances for pure RGB-based cuboid detection, and simulate random scenes with the objects for training and testing. In order to improve detection accuracy, all the simulated objects with limitless numbers and varied sizes can appear in high resolution images. As a baseline model, we validate our datasets using the YOLO-like standard deep learning architecture. In a coarse-to-fine manner, annotations such as cuboid, segmentation masks, and 3D models are first accessible on a down-sampled version of the simulated image and then on a sequence of higher resolution sub-images possibly utilized. Finally, we discuss experimental results and future directions.

Generators continue to deteriorate in performance due to aging and result in increased failure rates and reduced reliability. Therefore, studies are being conducted on anomaly prediction models for generator engines to prevent potential accidents during operation. However, there are problems in designing the models due to class imbalance and manual input of maintenance history. This study labels data from the time an anomaly occurs up to 60 minutes before the occurrence as anomalies to solve these problems. Data from the time an anomaly occurs up to 30 minutes before the occurrence were also added as derived variables to reflect the warning signs of anomalies in model training. The anomaly prediction models were created using engine log and maintenance history data and applying Random Forest(RF), eXtreme Gradient Boosting(XGB), Linear Support Vector Classifier(LSVC), and Deep Neural Networks(DNN) algorithms. The performance of the models was evaluated by F1-Score and Recall. XGB showed excellent performance in terms of F1-Score, and DNN in terms of Recall. As a result of comparing the F1-Scores to sort the optimal model for each system, XGB was optimal for systems 1, 2, and 4, and RF was optimal for systems 3 and 5. System 5 showed excellent performance when only the derived variable condition was applied, and the other systems showed excellent performance when applying the derived variable and labeling.

Nowadays, due to natural disasters the world is facing huge challenges such as economical, climatic, and losses a lot of precious human life. The traditional emergency response and rescue teams are physically visit different affected areas for inspection and save human lives. In this manual monitoring system created various problems such as human resources, time-consuming, and in real-time unable to accurately analyze the nature of the disaster. Therefore, there is an urgent need for an automatic real-time system to intelligently identified different disaster scenes and analyze the affected areas for quick response. Therefore, in this paper, an Unmanned Aerial Vehicles (UAVs) inspired framework is proposed for disaster scenes classification using a lightweight Convolution Neural Network (CNN). To validate the strength of the proposed framework a comparative analysis is conducted to show its superiority against different state-of-the-art models in terms of computational complexity and performance.

Real Time Location System (RTLS) refers to a system that provides various services by measuring location information of objects in real time. The RTLS system is being used in many fields related to the Internet of Things (IoT), such as medical, healthcare, performances, and production facilities. A high-accuracy positioning system is essential for quality of RTLS. A variety of methods such as triangulation, trilateration, and MDS are utilized for positioning, but each has its own drawbacks. We propose an efficient and accurate advanced positioning system with Deep Reinforcement Learning (DRL). In the learning environment, we adopt the Proximal Policy Optimization (PPO) algorithm and Adam Optimizer. The proposed system estimates the exact position with a small amount of computation using only distance information from four anchor nodes in a 3D environment. Through system performance evaluation, we proved that the proposed system showed superior performance compared to the existing system.

Knowledge graphs (KGs) play a pivotal role in modern applications such as decision-making systems, question answering systems, and searching and retrieval systems. However, the automatic construction of a knowledge graph from unstructured text is a challenging task. Moreover, traditional dictionary-, rule-based and supervised machine learning approaches are not reasonably practical due to their dependency on human-expert annotated resources. It is especially true when a knowledge graph is generated from domain-specific information, updated frequently, such as COVID-19 related information resources. This paper uses a pre-trained embedding model (BERT) to create word vectors from COVID-19 research articles. The proposed model is employed at two levels: entity extraction from the text and querying the knowledge stored in KG.

Fashion occupies a large part of the industry and has been a part of our lives. One of the ways to analyze trends in fashion is to detect and classify categories in fashion images. In this paper, we present fashion category detection through the utilization of Detectron2's Mask R-CNN, which is easy to learn with custom datasets and has a high model construction and learning speed. Learning is also done based on Fashionpedia, a large-scale fashion segmentation and attribute localization dataset built with fashion ontology. As a result, the average precision (AP) of the bounding box was 52.45 and that of segmentation was 48.77, showing reasonably high performances. We propose a possibility of using fashion category detection and classification work in the field of fashion design.

Blockchain has become the central research to enable data processing in edge computing. Considering that edge computing primarily deals with sensitive IoT data, privacy is an essential factor. Blockchain solves the secure data propagation problem among dynamically located edge nodes. However, blockchain enables public access for all connected nodes, thus leaving data privacy as an open issue. In this paper, we propose a smart-contract-based solution for controlling the data accesses in the public blockchain.

The reliability is an important issue for general systems. In IoT environments, the reliability is more important because an IoT application includes many IoT devices, gateways, and cloud servers. Applying a high overhead recovery mechanism to IoT devices that have resource constraints is difficult. In this paper, we designed a container-based checkpoint and restore system and ported the system to a wellknown open-source hardware platform (Raspberry Pi). By using this light-weight system, we can support seamless task handover between a fault robot and an alternative robot. We also implemented a testbed to verity the operability of the proposed system under a rescue robot scenario.

This paper proposes a hash index for spatiotemporal continuous query processing rules for filtering, classifying, analyzing, and responding to consecutively collected target objects. The Rete technique for improving the performance of rule-based complex event processing shows better performance than the rule interpretation method as it creates a compiled data structure for the rule. This paper proposes a performance improvement method that eliminates the rule search overhead by creating a spatiotemporal index for the Rete nodes expressing spatiotemporal continuity query rules and stabbing the Rete node of the rule by employing hash indexing on the stream data.

Vehicles with an Audio-Video Navigation (AVN) system have become another source of potential digital evidence. The AVN systems in modern cars retain information and event data from mobile devices connected to the car’s infotainment module, and navigation data in the form of tracklogs. These data provide a time history of a car’s geolocation that may be used to investigate an incident involving an automobile or reconstruct a crash. There has been little research into what types of user artifacts can be found on the AVN systems, and whether the AVN systems provide more vehicle event information than the connected mobile devices. For this study, we used the AVN system of a KIA NIRO EV vehicle for digital data acquisition and analysis. We have found that the AVN system provided some amounts of user data (start log, favorite routes, last destination search history, etc.). The acquired data and analyzed results can serve as basic digital evidence for vehicle forensic research.

Recently autonomous vehicles and connected cars are being equipped with various information and communication technology (ICT). The installed ICT on such vehicles and cars makes it possible to record events generated during communication between the built-in ICT module and the driver’s mobile phones and driver’s activity information. For example, the Audio Video Navigation (AVN) system, an ICT module, interacts with drivers and provides convenient features to drivers. AVN can be connected to a driver’s mobile device, and it can be used to store various driver-specific real time information such as driving records, phone call records, SMS, and music playback events. In this paper, we collect and analyze the digital data stored in an AVN installed in a KIA K5 sedan. Most of the analyzed data is related to the activities initiated from the mobile device connected to the AVN via Bluetooth. The result of our research reported in this paper can be useful for the forensic data analysis of vehicles.

The Internet of Things (IoT) technology is a recent development, and subsequently, various application services have appeared. Accordingly, diverse smart environments based on wireless networks are being developed. Home routers, which are widely used to build a wireless network environment at home, are exposed to many security threats. Therefore, setting up a secure router environment has become an important issue. In this paper, we propose an automated home router security configuration system (Auto- HRS) for multiple home routers located in a wireless network environment both by network administrators and those who install and operate their domestic routers at home. Auto-HRS analyzes the configuration HTTP request-response message of the home router and stores the HTTP message corresponding to the secure router configuration. Subsequently, the stored HTTP message is used to generate a secure-environmentsetting message and then send it to a home router to update the environment setting.

Stock costs in the securities exchange vacillate at each passing second; making it hard to anticipate the real stock value it will close at, at its end time. This leaves the financial backer in question about his benefit/misfortune edge against his speculation for that specific day. As APPLE Inc. is perhaps the most significant tech monster of the 21st century, many significant financial backers purchase APPLE (AAPL) stocks in the desire to make a fortune. Most existing arrangements utilize stock's end cost to decide its present market worth and utilize both Linear and Non-Linear Machine Learning models which appear to be incorrect. We adopted another strategy to decide APPLE Inc. stock cost by utilizing the OPENING worth of stock and foresee the stock's CLOSING value, which the financial backer acquires by the day's end and utilized Linear Regression model as it has ended up being quite possibly the most exact answers for stock forecast issues. Dataset was preprocessed, then, at that point prepared the model and, in the end, correlations show the genuine outcomes to approve the precision of my methodology and it end up being more than 95%. Results were practically equivalent to the current APPLE Inc. stock value which demonstrates the precision of my model.

Natural language processing, as an integral part of artificial intelligence technology, has foundations in a variety of disciplines, including linguistics, computer science, and mathematics. Rapid advances in natural language processing provide solid backing for machine translation research. This document first sets out the key concepts and key points of computational linguistics, followed by a brief review of the history and progress of NLP research in the United States and abroad. The document then summarizes the three stages of machine translation as well as the current state of research. Historically, the advancement curves of natural language processing and machine translation have almost coincided, as well as the two complement each other. On this premise, the paper examines NLP applications in machine translation and highlights problems and trends in the fields of artificial intelligence. Finally, the authors examine the link between machine translation and human interpretation in the era of artificial intelligence and speculate on machine translations long term prospects.

Gaussian process regression (GPR) is a nonparametric Bayesian methodology that is applied in various places in machine learning. GPR can identify uncertainty by learning data, predicting well, and obtaining variance in prediction. We conducted a study to predict and verify characteristics using a design parameter of a semiconductor using this GPR. In addition, by predicting the characteristic value of the secondary semiconductor derived from the predicted characteristics, it is possible to confirm the characteristics of the generated semiconductor.

Masonry structures account for a large proportion of the building stock worldwide. Presently, the structural conditions of such structures are mostly inspected manually, and which is expensive, laborious and subjective processes. As deep learning technique for computer vision advances, there is an opportunity to automate the visual inspection process using digital images. Several studies are in progress to automatically detect cracks in masonry structures using Deep Learning. However, it is important not only detecting a crack, but also measuring a length of the crack. This is because it is necessary to consider various factors required in the actual environment, such as calculating the cost of reinforcement work. In this paper, we propose the method that detects masonry cracks and measures the length of cracks with digital images. The aim of this study is to implement Deep Learning model for crack detection on masonry structure and to apply the method of crack length measurement additionally.

Transfer learning significantly improves the performance of a deep learning model on challenging datasets. However, the pre-trained models have certain constraints in terms of their architecture. For example, the state-of-the-art pre-trained models expect an input image with three-color channels because of the wide availability of color images. However, there are certain domains, e.g., medical applications, where grayscale images are produced and the models are required to perform certain tasks on them. Therefore, in this work we propose an approach to run pre-trained models on grayscale images while benefiting from transfer learning for multiclass classification task. We have used the MobileNetV2 pre-trained model to classify the CIFAR datasets. We have compared our results with a conventional method where the grayscale image is stacked up to form a pseudo-color image. Our analysis have shown that the proposed method reduces the computational time per epoch while improves the accuracy of the model.

The postal service sector uses machine learning to forecast delivery time and customer traffic. Studies on postal logistics forecasting have used various machine learning algorithms, but there were no attempts using Seasonal and Trend Decomposition using Loess (STL) decomposition, which is frequently used in other fields of time series forecasting. Therefore, this paper proposes a method of applying optimal STL decomposition cycles using the machine learning models of prior studies and the latest machine learning models. First, the proposed method decomposes the daily traffic using STL decomposition to generate three variables (Trend, Seasonal, and Residual). These variables are added to the existing input data variable to train the machine learning model. Finally, a suitable STL decomposition cycle for the model is selected to derive an optimal model. The proposed method was validated by creating nine machine learning (AdaBoost Regression, Random Forest Regression, Ridge, etc.) and two deep learning (DNN, LSTM) models and testing them. As a result, the application of STL decomposition reduced the forecast errors in all models except LSTM. In terms of the proposed method, linear regression had the lowest forecast error, and LSTM had the highest.

As there is no cure for fire blight, which mainly affects pears and apples, effective and rapid detection is very important. Existing fire blight diagnostic studies usually used biotechnology, such as immunodiagnostic kits. With the development of deep learning-based image recognition technology, an image-based fire blight diagnosis method has been proposed. For the diagnosis of diseases that have similar symptoms, including fire blight, this study developed a disease recognition model using the deep convolutional neural network (CNN). Fine-tuning was performed on VGG16, VGG19, ResNet50, DenseNet121, Inception-ResNet v2, NASNet and EfficientNet models, which were pre-trained through ImageNet dataset. The experiment used 14,304 images of six diseases collected from pear and apple as the dataset. As a result of the experiment, all seven fine-tuned models achieved an accuracy of more than 90%, among which the ResNet50 model achieved the highest accuracy at 98.83%. It is anticipated that the proposed model can be valuably used at actual farmhouses to diagnose and prevent fire blight through appropriate services in the future.

Recently, in computer vision behavior recognition is an active research area that plays a significant role in smart cities for crime prevention and urban safety. However, without base knowledge of Artificial Intelligence (AI) designing an efficient model is very difficult because we need data and programing skills for implementing the system. To tackle this problem, we designed and implemented a system that allows a user having no professional knowledge to easily and conveniently create a deep learning model. The interface of this system consists of Data Selection, Model Training and Testing, and Model Parameter values according to domains and categories. In addition, we designed a function to check the test results for the model selected by the user. This system allows users to quickly and easily create and test models.

Recently, for the collection and smooth management of big data, new value creation and efficient large-capacity data analysis technology by linking homogeneous and heterogeneous data platforms are required. In this paper, we designed a system that can easily link multiple data provided by homogeneous and heterogeneous platforms. The designed system uses REST-based Map API to construct a visualization environment, and we assume that it is a different DB by collecting and classifying public data from different institutions. We built our own server to connect and visualize different types of data through distance and address-based connections via the Haversine Formula, that is based on GPS coordinates by using image summary technology, heterogeneous data can be efficiently stored, analyzed, and visualized.

Covid-19 has been substantially impacting all major sectors of life since its outbreak in the early 2020. Owing to the sheer contagiousness and rapid transmission, the World Health Organization (WHO) issued stringent precautionary measures such as wearing facemask and keeping social distance to curb the spread of the pandemic. To enforce these precautionary measures, governments and multifarious private sectors across the world leveraged Deep Learning (DL) especially Computer Vision (CV). In this regard, the CV research community has paid greater focus on social distancing and facemask detection tools. DL undoubtedly exhibits better performance on large amount of properly annotated data. Therefore, this work focuses on the development of a large-scale and diversified facemask detection dataset that contains images of faces with masks and without masks under different lightning conditions and varying angles. The remarkable training and testing performance achieved by YOLOv4 on real-life test videos and movies, attests the diversity of the dataset samples.

Unmanned aerial vehicles (UAVs) or drones are versatile innovations that can capture pictures and videos and even collect air or soil samples. Natural disaster drones are especially critical, which help with understanding the damage after a disaster, locating people who need help, distributing resources and preparing for the next event. Computer vision, deep learning (DL), and drones can augment the existing sensors, thereby increasing the accuracy of natural disasters detector, and most importantly, allow people to take precautions, stay safe, and reduce the number of deaths and injuries that happens due to these disasters. Therefore, in this paper we propose a novel lightweight convolutional neural network (CNN) based framework to detect natural disasters including cyclone, flood, earthquake, and wildfire. The proposed CNN model is obtained by fine-tuning the MobileNetV2 that can be deployed on drones. Furthermore, the model is trained and evaluated using a publicly available natural disasters dataset by obtaining 83.4% accuracy. Similarly, the framework has ability to broad cast the notification in alarming situations, which makes our proposed framework a best fit for natural disasters detection in realworld surveillance settings.