Sakarya University Journal of Computer and Information Sciences

COVID-19 is a deadly virus that first appeared in late 2019 and spread rapidly around the world. Understanding and classifying computed tomography images (CT) is extremely important for the diagnosis of COVID-19. Many case classification studies face many problems, especially unbalanced and insufficient data. For this reason, deep learning methods have a great importance for the diagnosis of COVID-19. Therefore, we had the opportunity to study the architectures of NasNet-Mobile, DenseNet and Nasnet-Mobile+DenseNet with the dataset we have merged. The dataset we have merged for COVID-19 is divided into 3 separate classes: Normal, COVID-19, and Pneumonia. We obtained the accuracy 87.16%, 93.38% and 93.72% for the NasNet-Mobile, DenseNet and NasNet-Mobile+DenseNet architectures for the classification, respectively. The results once again demonstrate the importance of Deep Learning methods for the diagnosis of COVID-19.

The coronavirus epidemic, which began to affect the whole world in early 2020, has become the most talked about agenda item by individuals. Individuals announce their feelings and thoughts through various communication channels and receive news from what is happening around them. One of the most important channels of communication is Twitter. Individuals express their feelings and thoughts by interacting with the tweets posted. The aim of this study is to analyze the emotions of the comments made under the "daily coronavirus table" shared by the Republic of Turkey Ministry of Health and to measure their relationship with the daily number of cases and deaths. In the study, emotional classification of tweets was implemented using LSTM, GRU and BERT methods from deep learning algorithms, and the results of all three algorithms were compared with the daily number of cases and deaths.

The increase in environmental problems such as climate change and air pollution caused by global warming has risen the popularity of electric vehicles (EVs) used in the smart grid environment. The increasing number of EVs can affect the grid in terms of power loss and voltage bias by changing the existing demand profile. Effective predicting of EVs energy demand ensures reliability and robustness of grid use, as well as aiding investment planning and resource allocation for charging infrastructures. In this study, the electricity demand amounts in two different cities are modeled by Support Vector Regression, Random Forest, Gauss Process, and Multilayer Perceptron algorithms. The findings reveal that electric vehicle owners usually start to charge their vehicles during the daytime, the COVID-19 pandemic causes a serious decrease in EVs energy demand, and the support vector regression (SVR) is more successful in energy demand forecasting. Furthermore, the results indicate that the decrease in electricity demand during the COVID-19 pandemic caused reduces in the prediction accuracy of the SVR model (decrease of 17.1% in training and 12.6% in test performance, P

With the year 2020, the world faced a new threat that affects all areas of life, negatively affects production in all areas and paralyzes social life. The measures and restrictions taken by the country's governments to prevent the epidemic from spreading rapidly in the society with the effect of the coronavirus (Covid-19), which first appeared in China and spread all over the world, brought a new lifestyle. Covid-19 has been much the impact on electricity use and electricity production in the period in Turkey as like other countries. There was a sharp decline in commercial and industrial electricity use. The coronavirus effect has also been reflected in the electricity demand and the consumption amount has undergone a great negative change. Due to the enactment of measures against the new type of coronavirus epidemic and the partial or full-time curfews, electricity consumption was moved to homes, supermarkets and hospitals in April 2020 from places where mass consumption is intense, such as industry, workplaces and educational institutions. In this study, Covid-19 period, the first cases were examined electricity production and consumption in Turkey as of the date it is seen throughout, in comparison with electricity consumption data in the same month of the previous years corresponding to this period, the effects on electricity generation and consumption habits of this period were examined.

Geostationary Satellites (GEO) are being used commonly in communication market. The service providers uplink or downlink the signal by using their dedicated antennas (whether with or without tracking capability) to the GEO satellite. The satellite down-converts and amplifies the signal before sending back to the end users on Earth. Normally, the user set and adjust the uplink antenna to follow the GEO satellite movement as much as possible. As soon as there is no reduction in the link budget, this pointing assumed to be successful. On the other hand, the input power of the satellite, together with satellite longitude vs latitude, can give reasonable ideas about the accuracy of the ground antenna pointing. In the study, ground station pointing performance is shown with two different cases. One with tracking and one without tracking capability.

Celiac disease; is an autoimmune digestive system disease characterized by chronic intestinal inflammation and villus antrophy and triggered by dietary gluten genetically susceptible individuals. Diagnosis is based on serological tests and small bowel biopsy. Because of the diversity in the clinical features of the disease, various patient profile and the non-standardized serological tests, it is difficult to diagnose the celiac disease. Sensitivity, specificity, positive and negative predictive values are important parameters for the accuracy of the tests and they are missing in some clinicial studies. It is difficult do standardize the tests with these missing values for clinicians. The aim of this study is to train different machine learning algorithms and to test their performance in prediction of the diagnostic accurary parameters of celiac serological tests. Decision trees are effective machine learning algorithms for predicting potential covariates with %88,7 accuracy.

Augmented reality applications related with faces such as make-up, hair design, wearing glasses are mostly prepared for entertainment purposes. Facilitating the preparation of augmented reality applications and more accurate analysis of real-world data in applications will enable these applications to be used more widely in different sectors such as R&D, education and marketing. In generally, the steps in image-based augmented reality applications can be listed as follows; detection of the targeted object, finding two reference points for each targeted object in 2D images, determining the boundaries of virtual object in its image and inserting the virtual object in real time. In this study, the problems that may be encountered in preparations of these augmented reality applications expected to be used more in the future are examined through a case study. Firstly, haar cascade classifiers, used to find different face areas, are compared and as a result of the comparison, it is decided to use eye haar cascade. Afterwards, rule-based approaches have been used to eliminate the wrong ones among the found eyes and to match the eyes of the same face. Then the position, size and angle of the virtual object to be added are calculated and it is added to the face using affine transformations. The problems encountered in augmented reality and algorithms used for problem solving are explained through the virtual hat application, but these simply prepared algorithms, can be used for different objects such as hair and glasses by changing the target points.

In this study, we consider the solution of the road lighting problem by distributed algorithms over multi-agent networks where the objective is to determine the powers of the lamps that provide the desired road lighting level for a given road profile. The road is modeled as multiple road sections each with a length of 50 meters where a lighting pole is located in the middle of each section. Under given assumptions, the illumination levels of the road sections are expressed as linear functions of the powers of the lamps. When the processing units in the lighting poles can communicate wirelessly with the neighboring processing units and make simple calculations, it is shown that the power levels of the lamps that provide the desired lighting level for each road section can be calculated in a distributed manner. Finally, the model and the proposed solution has been verified by a numerical example.

In the academic field and as well as the application field, substantial attention has been drawn to coping with disasters. Since natural dangers causing a large proportion of disasters cannot be avoided, attempts to combat disasters have centered on preventing hazards from evolving into disasters through measures and restructuring works taken before, during, and after the disaster. There are many players involved in the disaster management process and many factors are influential in the effectiveness of this process. Among these factors, deciding the critical ones offers significant advantages, particularly in terms of practical studies. Concentrating on a single stakeholder in deciding the factors crucial to the success of this management structure, which has many stakeholders, can cause to ignoring the significant viewpoints of other stakeholder groups. Accordingly, for the evaluation of several success factors achieved as a result of a thorough and systematic literature review, the purpose of our study is to develop a common critical success factor model that will represent both the viewpoints of operational experts and academic experts, who constitute the stakeholders of this domain. Analytical Hierarchy Process (AHP) is utilized to determine the opinions of field experts while the text mining method was used to determine the perspectives of academics. In the study, therefore, a new AHP model assisted by text mining is introduced. Socio-cultural factors were brought to light by the analysis results of the suggested model. It has been determined by the results of the study that these two perspectives are overlapped largely in the organizational field and relatively in socio-cultural, environmental, and legal fields.

In recent years, different approaches and methods have been proposed to diagnose various diseases accurately. Since there are a variety of liver diseases, till late-stage liver disease and liver failure occur the symptoms tend to be specific for that illness. Therefore, early diagnosis can play a key role in preventing deaths from liver diseases. In this study, we compare the accuracy of different classification methods supported by the SAS software suite, such as Neural Network, Auto Neural, High Performance (HP) SVM, HP Forest, HP Tree (Decision Tree), and HP Neural for the diagnosis of liver diseases. In this study, the Indian Liver Patient Dataset (ILPD) provided by the University of California, Irvine (UCI) repository is used. Experimental results show that based on the metrics of our study, in the training phase while HP Forest achieves the highest accuracy rate, HP SVM and HP Tree do the lowest accuracy rates. However, in the validation phase, Neural Network achieves the highest accuracy rate and HP Forest does the lowest accuracy rate. Our experimental results may be useful for both researchers and practitioners working in related fields.

Büyük verilerin büyük hızlarla işlendiği çağımızda milyarlarca veriden farklı parametreler çıkararak çeşitli problemlerin çözümüne kolaylık getirmek için derin öğrenme algoritmaları kullanılmaktadır. Bu çalışmada, mevcut veri setlerinde bulunan kadın, erkek, yaşlı, genç, çocuk, bebek fotoğraflarının derin öğrenme algoritmaları ile cinsiyetlerini tespit etmek amaçlanmıştır. Bu tahminleme algoritmasını gerçekleştirmek için çeşitli derin öğrenme kütüphanelerinden faydalanılmış ve derin öğrenme modellerinden Alex Net ve VGG-16 ile yeni geliştirilen bir modelin diğer modellerle kıyaslanması yapılmıştır. Uygulamada kullanılan veri seti, kadın ve erkek fotoğraflarından oluşturulmuştur. Her fotoğraf ise kişi cinsiyetine ve yaşına göre etiketlendirilmiştir. Bu veri seti, 3170 eğitim verisi ile 318 test verisi içermektedir. Çalıştırılan üç farklı model sonuçları karşılaştırılmıştır. Makalede, derin öğrenme algoritmalarını kullanarak cinsiyet tahminiyapılması ayrıntılı bir şekilde incelenmiş ve yapılacak olan literatür çalışmalarına yol gösterilmesi, katkı sağlanması hedeflenmiştir.

Günümüzde birçok araç otonom olarak kontrol edilmek istenmektedir. Bu çalışmaların en önemli uygulama alanlarının başında askeri ve sağlık alanları olması münasebeti öncelik arz etmektedir. Bunun amaç için simülasyon ortamlarının geliştirilmesi hem zaman hem de ekonomik olarak daha elverişli araçlar ve imkanlar sunmaktadır. Buradan hareketle geliştirilen bir 3D sanal gerçeklik yazılımı sayesinde farklı kullanıcıdan alınan hem EEG hem de EMG sinyalleri paralel olarak MATLAB ortamına aktarılmış olup bu sinyaller sınıflandırılarak sanal oyuna komut olarak aktarılmış ve insansız hava aracı uzaktan yönlendirile bilinmiştir. Sistemin başarısını test etmek için farklı denekler üzerinde oluşturulan farklı rotalar kullanılarak performans analizleri yapılmıştır. Bu sayede donanımdan bağımsız olarak insandan elde edilen sinyaller ile sanal gerçeklik ortamı bütünleştirilmiş ve yapılan deneyler sonucunda başarılı bir şekilde kullanılabileceği sonucuna varılmıştır.

One of the languages available to describe a digital system in FPGA is the VHDL language. Since programming in hardware requires a different way of thinking than developing software, the students face some difficulties when trying to design in VHDL language with the previous and long experiences kept in mind in the learning of software imperative programming. These are its concurrency, parallel and sequential model. Due to the insufficient understanding of these topics, it is difficult for students to master the VHDL language. Analogies change the conceptual system of existing knowledge by linking the known to the unknown and by changing and strengthening their relationships. This study contributes to overcoming the problems that students encounter in the coding of the above-mentioned topics in VHDL language by using their experiences in traditional programming languages through analogies. Analogies were used in an undergraduate embedded systems course to explain complex concepts such as those related to signals, concurrent/parallel process; and to encourage comprehensive projects in digital circuit design. In feedback from students, the discussion and negotiation of analogies seems to minimize confusion and from using inappropriate expressions in using VHDL language.

Satellite operators utilize a two-stations turn around ranging (TAR) system to reduce the ground station measurement system's complexity and cost while having the same or better orbit determination accuracy for communication satellites orbit determination recently. This study investigates two stations' performance, four-way ranging on communication satellite orbit determination, operational conformance, and cost. The observation data sets are collected using traditional single station tracking (SST) and the new method TAR. The computed results using the Monte Carlo method encourage the satellite operators to use a four-way ranging system to observe and measure required data sets. TAR performance is evaluated, taking SST as a reference. The six classical orbital elements (a, e, i, RAAN, AoP, and TA) are compared for large numbers of observation data. The SST and TAR results are very close to each other. The worst-case calculated Euclidian distance between SST and TAR is 1.893 km at the epoch below the 6 km success criteria. The TAR observation method is appropriate to collect data sets for precise orbit determination. This work result indicates that satellite operators should consider deploying TAR stations to collect two-station range data sets and compute the orbit for nominal north-south station-keeping maneuvers (NSSK) and east-west station-keeping (EWSK) maneuver operations. The TAR method is superior to SST in terms of accuracy, operational conformance, and costs.

Deep learning, a subfield of machine learning, has proved its efficacy on a wide range of applications including but not limited to computer vision, text analysis and natural language processing, algorithm enhancement, computational biology, physical sciences, and medical diagnostics by producing results superior to the state-of-the-art approaches. When it comes to the implementation of deep neural networks, there exist various state-of-the-art platforms. Starting from this point of view, a qualitative and quantitative comparison of the state-of-the-art deep learning platforms is proposed in this study in order to shed light on which platform should be utilized for the implementations of deep neural networks. Two state-of-the-art deep learning platforms, namely, (i) Keras, and (ii) PyTorch were included in the comparison within this study. The deep learning platforms were quantitatively examined through the models based on three most popular deep neural networks, namely, (i) Feedforward Neural Network (FNN), (ii) Convolutional Neural Network (CNN), and (iii) Recurrent Neural Network (RNN). The models were evaluated on three evaluation metrics, namely, (i) training time, (ii) testing time, and (iii) prediction accuracy. According to the experimental results, while Keras provided the best performance for both FNNs and CNNs, PyTorch provided the best performance for RNNs expect for one evaluation metric, which was the testing time. This experimental study should help deep learning engineers and researchers to choose the most suitable platform for the implementations of their deep neural networks.

In this paper, the dataset of real incidents that occurred in Turkey between 2013 and 2017 and are regarded as acts of terrorism without any doubt according to Global Terrorism Database (GTD) are used to predict the group names responsible for unknown attacks. Principal Component Analysis (PCA) technique was used for feature selection. A novel voting method between five classification algorithms such as Random Forests, Logistic Regression, AdaBoost, Neural Network, and Support Vector Machine was used to predict the names. The results clearly demonstrate that the classification accuracy of all classifiers studied in this paper improved when PCA was used to select features as compared to selecting features without using PCA. The prediction of terrorist group names with PCA based feature reduction and the original features is carried out and the results are compared.

White blood cells (WBCs), which are part of the immune system, help our body fight infections and other diseases. Certain diseases can cause our body to produce fewer WBCs than it needs. For this reason, WBCs are of great importance in the field of medical imaging. Artificial intelligence-based computer systems can assist experts in the analysis of WBCs. In this study, an approach is proposed for the automatic classification of WBCs over five different classes using a pre-trained model. ResNet-50, VGG-19, and MobileNet-V3-Small pre-trained models were trained with ImageNet weights. In the training, validation, and testing processes of the models, a public dataset containing 16,633 images and not having an even class distribution was used. While the ResNet-50 model reached 98.79% accuracy, the VGG-19 model reached 98.19% accuracy, the MobileNet-V3-Small model reached the highest accuracy rate with 98.86%. When the predictions of the MobileNet-V3-Small model are examined, it is seen that it is not affected by class dominance and can classify even the least sampled class images in the dataset correctly. WBCs were classified with high accuracy using the proposed pre-trained deep learning models. Experts can effectively use the proposed approach in the process of analyzing WBCs.

The Internet of Things (IoT) refers to smart devices with limited resources that connect to the Internet and transmit data. Routing is an important process in this structure, which can be described as the general frame of wireless sensor networks (WSNs). The Routing Protocol for Low-Power and Lossy Networks (RPL) is recommended by the Internet Engineering Task Force (IETF) to provide communication in resource-constrained networks and is designed for routing in IoT. Basically, it is the Internet Protocol Version 6 (IPv6) protocol developed based on the energy consumed by devices. The algorithm has an important place in the performance of the IoT network. In this paper, the performance of the RPL under different objective functions (OFs) is examined. OFs are symbolized and defined by detailed equations. This study provides an experimental analysis of the RPL algorithm. An overview of the RPL algorithm is also included. Finally, the RPL algorithm is simulated by a custom simulator which is performing on the application layer, created using the Python programming language. The algorithm’s behaviour in terms of different OFs such as throughput maximization, energy efficiency maximization and energy consumption minimization was observed and the results were evaluated under different parameters such as packet size, number of nodes and different signal-to-noise ratio (SNR) values. Our experimental results may be useful for both researchers and practitioners working in related fields.

The next generation mobile networks have to provide high data rates, extremely low latency, and support high connection density. To meet these requirements, the number of base stations will have to increase and this increase will lead to an energy consumption issue. Therefore “green” approaches to the network operation will gain importance. Reducing the energy consumption of base stations is essential for going green and also it helps service providers to reduce operational expenses. However, achieving energy savings without degrading the quality of service is a huge challenge. In order to address this issue, we propose a machine learning based intelligent solution that also incorporates a network simulator. We develop a reinforcement-based learning model by using deep deterministic policy gradient algorithm. Our model update frequently the policy of network switches in a way that, packet be forwarded to base stations with an optimized power level. The policies taken by the network controller are evaluated with a network simulator to ensure the energy consumption reduction and quality of service balance. The reinforcement learning model allows us to constantly learn and adapt to the changing situations in the dynamic network environment, hence having a more robust and realistic intelligent network management policy set. Our results demonstrate that energy efficiency can be enhanced by 32% and 67% in dense and sparse scenarios, respectively.

Akülerde çalışma kalitesini belirleyen en temel ölçüt kutuplar arasındaki gerilim seviyesinin kararlılığı olup bu kararlılığı bozan temel sebep akü elektrolizinde meydana gelen "sülfatlaşma" olayıdır. Sülfatlaşma aküde plakalar arasında kısa devre oluşumu gibi akü açısından kritik arızalara sebebiyet verebilecek ve akünün servis süresinden daha kısa sürede değiştirilmesi ile sonuçlanacaktır. Bu çalışmada akülerde sülfatlaşma probleminin akü vazife görürken ve beklerken elektrokimyasal tepkime kontrolü yapılarak azaltılması veya kurşun asit akülerde kalıcı sülfatlaşmanın oluşumunun engellenmesi maksadıyla akü kutuplarından genliği ve frekansı ayarlanabilen sinyaller ile akünün uyarılarak elektrokimyasal tepkime kontrolü sağlayabilen bir akü koruma cihazı tasarımı yapılmıştır.

One of the important features of any real-time software is the worst-case execution time (WCET). To get an understanding of the timing behavior of real-time systems and to prove that the real-time software meets its deadlines, WCET analysis is performed. Today, researchers actively develop new WCET analysis methods and tools. Therefore, they need benchmark programs to evaluate and compare their work. To meet this need, in this study we present a new benchmark suite, called MBBench. MBBench includes a collection of C programs for Linux operating system and RTEMS real-time operating system. Its main aim is to help the evaluation and comparison of measurement-based WCET analysis methods/tools. MBBench has been published as open source. It can be obtained freely over the Internet.

Collaborative filtering algorithms are efficient tools for providing recommendations with reasonable accuracy performances to individuals. However, the previous research has realized that these algorithms are undesirably biased towards blockbuster items. i.e., both popular and highly-liked items, in their recommendations, resulting in recommendation lists dominated by such blockbuster items. As one most prominent types of collaborative filtering approaches, neighborhood-based algorithms aim to produce recommendations based on neighborhoods constructed based on similarities between users or items. Therefore, the utilized similarity function and the size of the neighborhoods are critical parameters on their recommendation performances. This study considers three well-known similarity functions, i.e., Pearson, Cosine, and Mean Squared Difference, and varying neighborhood sizes and observes how they affect the algorithms’ blockbuster bias and accuracy performances. The extensive experiments conducted on two benchmark data collections conclude that as the size of neighborhoods decreases, these algorithms generally become more vulnerable to blockbuster bias while their accuracy increases. The experimental works also show that using the Cosine metric is superior to other similarity functions in producing recommendations where blockbuster bias is treated more; however, it leads to having unqualified recommendations in terms of predictive accuracy as they are usually conflicting goals.

Sosyal yaşamın güvenliği açısından, suç işlenmeden önce, suçların önceden öngörülmesi ve gerekli önlemlerin alınması oldukça önemli bir konudur. Bu amaçla güvenlik birimlerinin gerekli önlemleri alması için suç analizi yapılması gerekmektedir. Bu konuda veri madenciliği yaklaşımı güvenlik birimlerine büyük verilerin analizinde önemli bir katkı sağlamaktadır. Bu kapsamda potansiyel suç bölgelerinin tahmin edilerek belirlenmesinde farklı veri analiz yöntemleri uygulanmaktadır. Suç bölgelerinin tespitinde ikili kümeleme yöntemlerini kullanarak suçun işlendiği bölgeler ile suç türlerini aynı anda kümelemek, geleneksel kümeleme yöntemlerine göre daha kapsamlı sonuçlar sağlamaktadır. Bu çalışmada veri madenciliği yaklaşımı ile suç bölgelerini belirlemek için "Boston'daki Suçlar” veri seti üzerinde CC ve Xmotif algoritmaları kullanılmıştır. Elde edilen ikili kümelerin etkinliğini ölçmek amacıyla algoritmaların performansı Chia ve Karuturi ikili küme skoruna (CKSB) bakılarak karşılaştırılmıştır. Elde edilen sonuçlar R-project 3.5.3 yazılımı kullanılarak sağlanmıştır. Kullanılan bu veri seti için CC algoritmasının Xmotif algoritmasına göre daha iyi sonuçlar verdiği ortaya çıkmıştır.

Machine learning and deep learning algorithms produce very different results with different examples of their hyperparameters. Algorithm parameters require optimization because they aren't specific for all problems. In this paper Long Short-Term Memory (LSTM), eight different hyperparameters (go-backward, epoch, batch size, dropout, activation function, optimizer, learning rate and, number of layers) were used to examine to daily and hourly Bitcoin datasets. The effects of each parameter on the daily dataset on the results were evaluated and explained These parameters were examined with hparam properties of Tensorboard. As a result, it was seen that examining all combinations of parameters with hparam produced the best test Mean Square Error (MSE) values with hourly dataset 0.000043633 and daily dataset 0.00073843. Both datasets produced better results with the tanh activation function. Finally, when the results are interpreted, the daily dataset produces better results with a small learning rate and small dropout values, whereas the hourly dataset produces better results with a large learning rate and large dropout values.

Parabolic blending (PB) is one of the important topics in applied mathematics and computer graphics. The use of generalized parabolic blending (GPB) for different scenarios adds flexibility to the polynomial. Overhauser (OVR) elements is a special case in GPB (r=0.5, s=0.5). GPB can also be used in estimation. In this study, data obtained from thickness distribution of a 3mm thick high impact polystyrene product after thermoforming using a mold was used for data estimation. For this purpose, software has been developed. The software development steps and formula usages are explained. Using the developed software, polynomials for GPB and default PB (OVR) were created. The data set was compared with the y values produced by the polynomials for certain x values. At the end of the research, it was determined that the results obtained from the GPB were 0.1728 percent more accurate than the data obtained from the PB for the default values.

The development and widespread use of computer systems has increased the need for secure storage of data. At the same time, the analysis of digital data storage devices is very important for forensic IT professionals who aim to access information to clarify the crime. File systems of disk drives use partition structures to securely store data and prevent problems such as corruption. In this study, deletion or corruption of partitions on commonly used DOS/Master Boot Record (MBR) configured hard disk drives are investigated by using forensic tools. In order to analyze hard disk drives, Forensic Tool Kit (FTK), Magnet AXIOM, Encase, Autopsy and The Sleuth Kit (TSK), which are widely used as commercial and open source, are analyzed by using a presented scenario. In the scenario, the primary partition and the extended partition are created using the DOS/MBR partitioning structure on the test disk. Test files are added to the sections and the sections are deleted. The digital forensics tools were tested on the presented scenario. According to the obtained results, TSK and Encase are successful tools for DOS/MBR structured HDD analysis. However, FTK, Magnet AXIOM and Autopsy could not achieve information detection on DOS/MBR structured disks. These results clearly demonstrated that crime data can be hidden in MBR structured HDD. To carve these data, the correct methodology should be selected.

In this paper we present effect of the chaotic crossover operator with different chaotic maps on the metaheuristic search algorithm Breeding Swarms algorithm which is the Particle Swarm Optimization’s one of the genetic algorithm hybrid form. Some of the many optimization problems could have too many local extrema. Most of the time optimization algorithms could stuck on these extrema therefore these algorithms could have trouble with finding global extremum. To avoiding local extrema and conduct better search on search space, a chaotic number generator is used on Breeding Swarms algorithm’s most of the random procedures. To test efficiency and randomness of the chaotic crossover operator, different chaotic maps are used on the Breeding Swarm algorithm. Test and performance evaluations are conducted on Multimodal and unimodal benchmark functions. This new approach showed us that modified Breeding Swarms algorithm yielded slightly better results than Particle Swarm Optimization and original Breeding Swarms algorithms on tested benchmark functions.

Electric energy passes through many stages from production to usage, and many electrical connections take place at these stages. Electrical connections are quite important for system safety and electrical energy quality. In this study, current-temperature relationship in the cable was investigated in cable joints made with screw connection terminals, depending on torque applied to screw. In numerical solution of the problem, Comsol Multiphysics program based on Finite Element Method was used first, the screw-conductor interface was investigated based on applied torque, and then electrical-thermal analysis was performed on this geometry. Experimental studies were carried out to demonstrate accuracy of digital model, and conductors and terminals used in household installations were taken into account in these studies. Amount of deformation of conductors depending on torque applied to screw and its effect to current carrying capacity of these terminals having screw connection were investigated. In this study, the appropriate torque value is 0.4 Nm, and the maximum temperature value is 45 oC on contact surface. Also, it has been shown that the optimum torque value to be applied in the copper conductor cable joint with 1.5 mm2 cross section area is 0.4 Nm.

Capital investment cost is the major obstacle to the increasing share of electricity from renewable energy sources (RES-E). Therefore, RES-E incentive mechanisms are incorporated into markets to compensate cost-related barriers and to increase RES-E deployment rate. In this study, the impact of direct capital investment subsidy on RES-E in generation expansion planning (GEP) has been analyzed and deployment rates of renewable power plants have been defined. The effect of current subsidy mechanisms on the installed power capacity of various sources has also been analyzed and policy recommendations have been put forth in the light of the characteristics of Turkey’s current subsidization mechanism and its outcomes. Genetic algorithm was applied to solve the GEP problem. The share of non-hydro renewable power plants for future additions in overall installed power was determined as 9.45% without the proposed incentive, while it was estimated to rise to 13.65% when it was promoted by direct capital investment subsidy of 50%. The deployment rates of renewable power plants are expected to grow as the imported coal share in total installed power is expected to decline after applying the proposed subsidy.

Expressing the visual content of an image in natural language form has gained relevance due to technological and algorithmic advances together with improved computational processing capacity. Many smartphone applications for image captioning have been developed recently as built-in cameras provide advantages of easy-operation and portability, resulting in capturing an image whenever or wherever needed. Here, an encoder-decoder framework based new image captioning approach with a multi-layer gated recurrent unit is proposed. The Inception-v3 convolutional neural network is employed in the encoder due to its capability of more feature extraction from small regions. The proposed recurrent neural network-based decoder utilizes these features in the multi-layer gated recurrent unit to produce a natural language expression word-by-word. Experimental evaluations on the MSCOCO dataset demonstrate that our proposed approach has the advantage over existing approaches consistently across different evaluation metrics. With the integration of the proposed approach to our custom-designed Android application, named “VirtualEye+”, it has great potential to implement image captioning in daily routine.

For studying operational and organizational systems, modeling and simulation tools are becoming increasingly relevant. It is possible to build a lot of systems and study their behavior, which saves a lot of effort, time, and cost, where it cannot or difficult to study its behavior in the real world. There are many frameworks to implement modeling and simulation using a computer. in this paper DEVS-Suite for discrete events is used to implement a simulation of cargo unloading problem which represents a study on estimating the optimal number of trucks and cranes required in process of unloading goods and according to some determinants. The duration of the simulation is one month which is equivalent to 43,200 minutes. Based on the performance measures that were adopted in this study, the optimal number of trucks and cranes is 5 out of three assumptions of 3, 5, and 10, where the work will be in a permanent state of work and high productivity.

Nowadays, chaotic systems are used quite often, especially in image encryption applications. Hypersensitivity to the initial conditions, limited field-changing signs and irregular movements make these systems one of the critical elements in scientific matters such as cryptography. These systems are divided in two parts as discrete time and continuous time in terms of their dimensions and properties. Gray level image encryption applications generally use one-dimensional and color image encryption applications generally use multi-dimensional chaotic systems. In this study, Tent Map, Cat Map, Lorenz, Chua, Lu chaotic systems were used for chaotic neural network based image coding application and Logistic Map and 3D Cat Map chaotic systems were used for 3D chaotic Cat Map based image encryption application. The encrypted image and the original image were examined by various analysis methods. As a result of these analyzes, it was seen that both applications gave very successful results. Analyzes have shown that the chaotic neural network based image coding algorithm is more secure and successful.

Urine sediment tests are important in the diagnosis of abnormal diseases related to the urinary tract. The formation of cells such as red blood cells and white blood cells in the urine of patients is important for the diagnosis of the disease. Therefore, cells need to be fully identified in clinical urinalysis. Urinalysis with human eyes; Since it is subjective, time consuming and causing errors, methods have been developed to automate microscopic analysis with the help of computer and software systems. In this study, the YOLO-v7 algorithm, which gives successful results in image processing technology, was used as a method and model. The dataset used in the study was created by using microscopic images of urine sediment taken from the Biochemistry Laboratory of the Faculty of Medicine, Selcuk University. Seven different cell segmentation and classification studies have been carried out, including WBC, RBC, WBCC, Epithelial, Flat Epithelial, Mucs and Bubbles, which have clinical value for the diagnosis of the disease. Experimental studies were carried out with the YOLO-v7 algorithm and the results were presented. The contributions of this study can be summarized as follows. (1) In this study, which is proposed for segmentation of cells on the urine cell images in the Urine Sediment dataset, for the experimental studies carried out with the YOLO model, whose performance was evaluated; Precision, Recall, mAP(0.5) and F1-Score(%) segmentation performance metrics were calculated as 0.384, 0.759, 0.432 and 0.510, respectively. (2) A computer-aided support system to assist physicians in segmenting urine cells is presented as a secondary tool. Classification accuracy for WBC, RBC, WBCC, Epithelial, Flat Epithelial, Mucs and Bubbles cells was calculated as 0.78, 0.94, 0.90, 0.57, 0.92, 0.68 and 0.97, respectively. A mean classification success of 0.822 was achieved for all classes. Thus, it has been seen that the Yolov7 model can be used by experts as a tool for recognizing cells in the urine sediment. As a result, it has been shown that suitable deep learning models can be used to recognize the biometric properties of urinary sediment cells. With the model created using deep learning libraries, urine sediment cells can be easily classified, and it is possible to define many different cells if there is a dataset with sufficient number of images.

With urbanization, population, and consumption on the rise, urban waste generation is steadily increasing. Consequently, waste management systems have become integral to city life, playing a critical role in resource efficiency and environmental protection. Inadequate waste management systems can adversely affect the environment, human health, and the economy. Accurate and rapid automatic waste classification poses a significant challenge in recycling. Deep learning models have achieved successful image classification in various fields recently. However, the optimal determination of many hyperparameters is crucial in these models. In this study, we developed a deep learning model that achieves the best classification performance by optimizing the depth, width, and other hyperparameters. Our six-layer Convolutional Neural Network (CNN) model with the lowest depth and width produced a successful result with an accuracy value of 89% and an F1 score of 88%. Moreover, several state-of-the-art CNN models such as VGG19, DenseNet169, ResNet101, Xception, InceptionV3, RegnetX008, RegnetY008, EfficientNetV2S trained with transfer learning and fine-tuning. Extensive experimental work has been done to find the optimal hyperparameters with GridSearch. Our most comprehensive DenseNet169 model, which we trained with fine-tuning, provided an accuracy value of 96.42% and an F1 score of 96%. These models can be successfully used in a variety of waste classification automation.

Pneumonia is a general public health problem. It is an important risk factor, especially for children under 5 years old and people aged 65 and older. Fortunately, it is a treatable disease when diagnosed in the early phase. The most common diagnostic method known for the disease is chest X-Rays. However, the disease can be confused with different disorders in the lungs or its variants by experts. In this context, computer-aided diagnostic systems are necessary to provide a second opinion to experts. Convolutional neural networks are a subfield in deep learning and they have demonstrated success in solving many medical problems. In this paper, Xception which is a convolutional neural network was trained with the transfer learning method to detect viral pneumonia, normal cases, and bacterial pneumonia in chest X-Rays. Then, five different machine learning classification algorithms were trained with the features obtained by the trained convolutional neural network. The classification performances of the algorithms were compared. According to the test results, Xception achieved the best classification result with an accuracy of 89.74%. On the other hand, SVM achieved the closest classification performance to the convolutional neural network model with 89.58% accuracy.

Image hashing is an algorithm used to represent an image with a unique value. Hashing methods, which are generally developed to search for similar examples of an image, have gained a new dimension with the use of deep network structures and better results have started to be obtained with the methods. The developed deep network models generally consider hash functions independently and do not take into account the correlation between them. In addition, most of the existing data-dependent hashing methods use pairwise/triplet similarity metrics that capture data relationships from a local perspective. In this study, the Central similarity metric, which can achieve better results, is adapted to the deep reinforcement learning method with sequential learning strategy, and successful results are obtained in learning binary hash codes. By taking into account the errors of previous hash functions in the deep reinforcement learning strategy, a new model is presented that performs interrelated and central similarity based learning.

Elektromanyetik fırlatıcılarda merminin üzerindeki kuvvet, uyartım değeri ve merminin sargı içerisindeki konumuna göre değişiklik göstermektedir. Bu çalışmada elektromanyetik fırlatıcılarda kullanılan bobin ve merminin 3 boyutlu modeli oluşturularak sonlu elemanlar metodu ile analizler gerçekleştirilmiştir. Parametrik çözüm metodu kullanılarak, sargının uyartım değeri ve mermi konumu değiştirilerek mermi üzerindeki kuvvet karakteristiği elde edilmiştir. Sonlu elemanlar analizlerinde daha küçük çözüm adımları tanımlanarak daha hassas analizler gerçekleştirilebilir. Bununla birlikte, değişkenlerin sayısındaki artış nedeniyle analiz süresi uzamaktadır. Analiz süresi dikkate alınarak, çalışmada kuvvet kestirimi tek gizli katmandan ve iki gizli katmandan oluşan çok katmanlı sinir ağı modelleri kullanılarak gerçekleştirilmiştir. Çok katmanlı sinir ağları ile yapılan kuvvet kestirimi çalışmalarında başarılı sonuçlar elde edilmiştir.

This paper introduces a new measure for individual thermal comfort, inspired by the current standards for population thermal comfort, and a statistical model allowing us to imitate individuals’ thermal comfort preferences. Our approach is based on the observation that an individual has a temperature range around his or her desired temperature point in which he or she is comfortable with the surrounding thermal environment. The crucial parameters of our statistical model, which represents the thermal characteristic of individuals of building occupants, have been assumed to be normally distributed random variables so that the thermal comfort preferences of different individuals can be generated for the further simulation purposes. When aggregated to a population’s general thermal comfort parameters, the variables of these distributions have been adjusted in such a way as to bring very close consistency with the current standards, which define the criteria for acceptable thermal conditions of human occupancy in a built environment.

Malaria is a disease caused by a parasite. The parasite is transmitted to humans through the bite of infected mosquitoes. Thousands of people die every year due to malaria. When this disease is diagnosed early, it can be fully treated with medication. Diagnosis of malaria can be made according to the presence of parasites in the blood taken from the patient. In this study, malaria detection and diagnosis study were performed using The Malaria dataset containing a total of 27,558 cell images with samples of equally parasitized and uninfected cells from thin blood smear slide images of segmented cells. It is possible to detect malaria from microscopic blood smear images via modern deep learning techniques. In this study, 5 of the popular convolutional neural network architectures for malaria detection from cell images were retrained to find the best combination of architecture and learning algorithm. AlexNet, GoogLeNet, ResNet-50, MobileNet-v2, VGG-16 architectures from pre-trained networks were used, their hyperparameters were adjusted and their performances were compared. In this study, a maximum 96.53% accuracy rate was achieved with MobileNet-v2 architecture using the adam learning algorithm

For Turkey to achieve the targets of Vision 2023 of being in the top ten economies of the world, the eleventh National Development Plan (NDP11) focuses on ensuring uninterrupted, high-quality, sustainable, reliable and affordable energy supply. In this regard medium- and long-term energy supply-demand planning is regarded as a key input to the planning process. Medium and long-term planning is possible only when reliable forecasts are available. Using Turkey’s electrical energy consumption data from 1970 to 2015, this study presents novel Gaussian, Fourier and Exponential curve fitting and extrapolation approaches to forecast Turkey’s electrical energy consumption up to the year 2025. Major interest is put on how the model forecasts electrical energy consumption for year 2023 because this year marks a century of the establishment of the Republic of Turkey and all strategic plans are focused on how to achieve the targets as outline in Vision 2023. We evaluate the performance of the models on how best they forecast electrical energy consumption for the year 2023. Our forecasts for the year 2023 are 352.7TWh, 377.4 TWh, and 460.1TWh for the Gaussian, Fourier and Exponential models respectively which compare well with NDP11’s estimated 375.8 TWh electrical energy consumption in 2023.

This study addresses a novel perspective for analyzing the source of confidence in human behavior. The concept of confidence was examined via the relationship between two phenomena in the area of machine learning and psychology, namely the Dunning-Kruger effect and the curse of dimensionality. A relationship was established between these two phenomena which were investigated in the light of neuroscience. This study claims that confidence is highly related with the total time it takes to reach specific information and this relationship is inversely proportional. Image gender classification algorithm was used to analyze this relationship for this study and the curves which were obtained as a result of this analysis was compared with the curve of Dunning-Kruger effect and curse of dimensionality. This relationship has been explained by the knowledge of human's problem-solving ability and mathematical models of memory.

The main purpose of this work is to examine the environmental impact of a daylight-linked dimming lighting control system integrated in the Lighting Laboratory of Electrical and Electronics Engineering Department, Sakarya University. For this purpose, total annual energy savings and greenhouse gas emission savings is performed in terms of measured annual in operation data and calculated life cycle energy data. The results indicate that the system provides 1,519.55 kWh annual energy savings and spends 365.26 kWh life cycle energy. Assuming that life time of a lighting control system is ten years, annual energy spent by the control system is estimated 36.54 kWh/year. Total annual lighting energy savings, subtraction of estimated annual life cycle energy from measured annual energy savings, are calculated 1,483.01 kWh which is nearly 40% of total annual lighting energy consumption of the test room accordingly. In conclusion, it is established that emissions of the test room are reduced 2.71 tCO2 annually by the lighting control system proposed in this work.

Bu çalışmada, 2009-2019 yılları arasında Tıpta derin öğrenme ile ilgili yapılmış çalışmalar, derin öğrenmenin Tıbbı görüntü, veri ve sinyaller üzerine başarısını gözlemlemek için araştırılmıştır. Web of Science’tan elde edilen çalışmalar değerlendirilmiş ve atıf sayısına göre seçilmişlerdir. Çalışmalar yayın yılı, derin ağ yapısı, kullanılan veritabanı ve değerlendirme kriterine göre tablo haline getirilmiştir. The results have shown that the deep learning network structures, applied on fundus images, have attained nearly %99 percent accuracy. Sonuçlar retinal fundus görüntüleri uygulanan derin öğrenme ağ yapılarının doğruluklarının %99’lara ulaştığını göstemektedir. Bu aralıktaki çalışmaların çoğu radyoloji ve nükleer tıp alanında yapılmış olsa de sonuçlar henüz %80-90 aralığında görülmektedir. Bu sonuçlar bilgisayar destekli teşhis sistemlerinin çok yakın bir gelecekte tam performans ile kullanılacağını göstermektedir.

Image processing techniques give highly successful results when used deep learning in classification studies. Applications benefit from this kind of work to make life easier. In this study, a mobile application is developed that takes photo of a plant and makes image processing on it to provide information about its name, the time to change the soil, the amount of sun light and nutrition it needs. The model is trained using the Convolutional Neural Networks, and dataset is successfully applied to the network. Currently, the application is capable to classify 43 different plants in mobile environment, and its classification capacity is planned to be expanded with new plant species as a future study. Up to 90% accuracy is reached in this study with the current version of the application.

Osteoporoz, vücudumuzdaki kemiklerin sertliklerinin azalıp, kalitelerinin bozulması sonucunda daha zayıf ve kırılabilir hale gelmeleri ile ortaya çıkan ve tüm iskeletimizi etkileyen sistemik bir hastalıktır. Bu çalışmada, bir iskelet hastalığı olan osteoporozun ön tanısında kullanılan X-ray absorbsiyometri (DEXA) testinin radyasyon dezavantajı sebebiyle, buna alternatif ve yapay zeka tabanlı, doğruluk değeri yüksek bir karar destek sistemi oluşturmak amaçlanmıştır. Gerçekleştirilecek sistem bir ön tanı yöntemi olarak kullanılacaktır. Bunun için, 70 hastadan alınan belirli parametrelerden oluşturulan veri seti yardımı ile tasarlanan olasılıksal sinir ağı (OSA) kullanılmıştır. Elde edilen başarı oranı ile Yapay sinir ağlarının osteoporoz hastalığının teşhisinde karar destek sistemi olarak kullanılabileceği görülmüştür. Bu çalışma sayesinde bu hastalığın şüphesi ile ilgili birime gelecek tüm hastalara DEXA testinin uygulanma olasılığı aza indirgenmiş olacaktır.