Search

Oyungeliştirmenin en büyük zorluklarından biri, hem tatmin edici derecede gerçekçihareket sonuçları üreten hem de oyun geliştiricilerin sınırsız hayal gücününyarattığı dünyalarda farklı senaryoları gerçekleştirebilen bir yol bulmaalgoritması üretmektir. Ayrıca oyunlar son kullanıcıya yönelik programlarolduğu için içerdikleri sistemlerin mümkün olduğunca az bilgisayar kaynağıkullanması ve maliyet açısından mümkün olan en kısa sürede geliştirilmesiistenmektedir. Mevcut yol bulma çözümleri, soruna güçlü yanıtlar üretebilse deçözülmesi uzun bir geliştirme süresi gerektiren bazı kronik sorunları daiçermektedir. Bu çözümler, baştan sona gezinilebilen ve sürekli haritalarda çokiyi çalışmaktadır. Ancak yürüme dışında zıplama, uçma gibi farklı hareketmekaniklerinin de kullanıldığı çeşitli engellerin aşılması gereken durumlaraçözüm getirememektedir. Çoğu zaman geliştiricilerin yol bulma örgülerinibirbirine linklerle manuel olarak bağlaması gerekmektedir. Bu da farklı hareketmekaniklerine sahip haritalarda oyun geliştirme sürecini önemli ölçüde uzatmaktadır.Aynı zamanda bağlanan linkler manuel olarak kurulduğu için link üzerindehareket eden cismin hareketi doğal görünmemektedir. Bu çalışmanın odak noktası,mevcut yol bulma algoritmalarının zorluklarını aşmak için yapay sinir ağları vederin pekiştirmeli öğrenme kullanarak bir düğüm ağı oluşturacak bir sistemoluşturmaktır. Son olarak, yapım aşamasında yapay sinir ağları kullanılmayacağındanson kullanıcı için hızlı ve daha az kaynak kullanan bir sistemhedeflenmektedir. One ofthe biggest challenges of game development is to produce a pathfindingalgorithm that both produces satisfactory realistic movement results and cansolve different scenarios in worlds created by game developers' unlimitedimagination. Furthermore, since the games are programs for the end user, it isdesired that the systems they contain use as little computer resources aspossible and be developed as quickly as possible in terms of cost. Althoughexisting solutions can produce strong answers to the problem, they also containsome chronic problems that take a long development time to solve. Existing solutionswork very well on maps that are continuous and can be navigated by moving fromstart to finish. However, they cannot find a solution in cases where variousobstacles must be overcome by various movement mechanics such as jumping,flying are used other than walking. Most of the time developers must manuallyassign links to meshes, significantly prolonging the game development processon the maps with such features. At the same time, since the connected links areestablished manually, the movement of the object moving on the link does notseem natural. The focus of this study is to create a system that will generatea node network by using artificial neural networks and deep reinforcementlearning to overcome the difficulties of existing pathfinding algorithms.Finally, a system that is fast and uses less resources is aimed for the enduser, since artificial neural networks will not be used during the build phase.&nbsp

One of the biggest challenges of game development is to produce a pathfinding algorithm that bothproduces satisfactory realistic movement results and can solve different scenarios in worlds created by gamedevelopers' unlimited imagination. Furthermore, since the games are programs for the end user, it is desired thatthe systems they contain use as little computer resources as possible and be developed as quickly as possible interms of cost. With the advances in software, lightweight methods have been developed that can provide generalmap coverage in this regard. Although existing solutions can produce strong answers to the problem, they alsocontain some chronic problems that take a long development time to solve. Existing solutions work very well onmaps that are continuous and can be navigated by moving from start to finish. However, it cannot find a solutionin cases where various obstacles must be overcome by various movement mechanics such as jumping, flying ordash are used other than walking. Developers must manually assign links to meshes, significantly prolonging thegame development process on the maps with such features. At the same time, since the connected links areestablished manually, the movement of the object moving on the link does not seem natural. The focus of thisstudy is to create a system that will generate a node network by using artificial neural networks and deepreinforcement learning to overcome the difficulties of existing pathfinding algorithms. Finally, a system that isfast and uses less resources is aimed for the end user, since artificial neural networks will not be used during thebuild phase. Some promising results have been obtained so far, which shows that the alternative methodologyproposed in this study could be a useful alternative for game developers.

This paper aims to provide an analysis of anaphoric relations in tweets in Turkish language. The analysis offered rests on the results of a sequence of experiments conducted using a group of machine learning algorithms. The algorithms used in this study are J48, Voted Perceptron, SVM (support vector machine), Naive Bayes and k-nearest neighbours. These classifiers were experimented by parametric variations are scrutinized to elaborate on the problem of matching a model conveniently to the task available already. Another important contribution of the paper is the comparison offered between two genres of texts, namely tweets versus child stories. Our experimental results are compared with those of the previous work and, thereby, a comparison is offered between the anaphoric structure of tweets and that of child stories in Turkish.

The purpose of a literature review is to gain an understanding of the existing research and debates relevant to a particular topic or area of study and to present that knowledge in the form of a written document. A literature review should identify critical knowledge gaps and thus motivate researchers to close this breach. is, writing a review not only requires an examination of past research but means making a chart for research. It has been proven that literature review is essential and highly critical for the success of any scientific research. It has also been proven that a literature review should not just be a summary of each source. Instead, it requires the comparison of each source to other relevant literature on the topic, the evaluation of each of the previous studies, elaboration of each reference’s contribution to the body of knowledge, and the integration of the sources into the proposed argument. There are several methodologies and models to carry out a literature review, such as concept-centric review, author-centric review, or a mix of both, using tables to distinguish and summarize the topics, meta-studies, thematic analysis, and other qualitative and quantitative approaches. In some literature reviews, even descriptive statistics are used to summarize the common properties or concepts in the relevant studies. However, to the best of our knowledge, no data mining approach has been used in literature reviews. This study proposes a novel approach for the quantitative analysis of literature reviews with the aid of association rules. The features are decided upon and extracted from the relevant studies in a specific literature review. Then, these features are transformed into nominal attributes and a dataset that could be used for association rule mining. Apriori algorithm, which is the most common algorithm in association rule mining is applied to this dataset and the resultant rules are filtered and analyzed. This new approach has been applied to the literature survey of a published Ph.D. dissertation and some promising results have been obtained. Some of the rules obtained by this novel approach conform to the deductions and conclusions of the literature review in the Ph.D. thesis. This study shows that the use of association rules could be an effective and supportive approach for the analysis of literature reviews for social and/or natural sciences.

Coal, as most common fossil-fuel, has been subjected to several geochemical and mineralogical studies due to presence of potentially toxic elements. In these studies, statistical methods (e.g., correlation coefficient, cluster and factor analyses) are mainly used for determination of toxic elements affinities in coal. In environmental concerns, the statistical methods are recently topic of discussion due to correlations between some elements with minerals that cannot be chemically affiliated. Nevertheless, microanalyses methods as like scanning electron microscopy-energy dispersive spectrometer (SEM-EDX) or electron microprobe (EPMA) and machine learning algorithms more commonly applied in determination of affinities of some toxic elements in coal. This study aims to correlate geochemical and mineralogical data of lower (kM2) coal seam in the Soma coalfield with Bray-Curtis, Cosine and Tanimoto similarities and different similarity measures like Pearson correlation co-efficiencies. The results of similarity measures evaluated using agglomerative hierarchical clustering algorithm (average linkage) and elements grouped in several clusters. Most of identified elemental groups, expected a few of them, based on Canberra, Chebyshev, Bray-Curtis and Tanimoto measures do not appear to be in agreement with mineralogical and geochemical data. Nevertheless, elements affiliated with aluminosilicate elements (e.g., Al, K, B, and Cs) are grouped in together, and elements (e.g., S, As, Mo and U) related with redox conditions in coal formation environment are located in the same group according to Pearson correlation co-efficiencies and cosine similarity. In addition, this data is in agreement with SEM-EDX and XRD data of studied coal samples. These results imply that cosine similarity could be an alternative for Pearson correlation coefficiency’s method in coal studies. As a result, more detailed studies using both similarity measures should be conducted in future, and these similarity measures should always be correlated with SEM-EDX and XRD data.

The hierarchical clustering algorithm, especially Pearson correlation coefficient, along with other statistical approach, is used statistical approach for determining the toxic elements affinities in coal, and this method is one of the common approaches due to the correlations between some elements with minerals that cannot be chemically affiliated. This study aims to correlate geochemical and mineralogical and data of the lower seam (up to 30 m) in the Soma coalfield with Cosine and Bray-Curtis similarity measures, Chebyshev and Canberra distance metrics, and also Pearson correlation and Tanimoto coefficients. The results have been evaluated using agglomerative hierarchical clustering algorithm with average linkage, and elements are grouped in several clusters. Some of the similarity measures and distance metrics do not seem to agree with mineralogical and geochemical data. However, elements affiliated with aluminosilicate elements (e.g., Al, K, and Cs) are grouped, and elements (e.g., As, Mo and U) related to redox conditions in coal depositional environment are located in the same group according to Pearson correlation coefficient and Cosine similarity. In addition, this data has been observed in agreement with SEM-EDX and XRD data of studied coal samples. The present study indicates that Cosine similarity could be an alternative for the Pearson correlation coefficient in coal studies.

Kötücül yazılımların tespiti Windows, Mac, Linux gibi geniş kitlelerin kullandığı işletim sistemleri de dahil olmak üzere, platformdan bağımsız bir biçimde karşımıza çıkmaktadır. Android işletim sistemi, akıllı telefonlarda pazar payı liderliği ve açık kaynaklı yapısıyla kötücül yazılımların birincil hedefi haline gelmiştir. Bunun sonucu olarak da bilişim suçlularının günümüzde öncelikli hedefi olan bu platform, aynı zamanda yeni güvenlik yöntemleri ve teknikleri tasarlayıp yeterliliklerini ölçmek isteyen araştırmacılar için de en öncelikli ortamlardan birini oluşturmaktadır. Olabildiğince fazla sayıda ve güncel çalışmaların incelenmesinin amaçlandığı bu literatür taramasında, Android işletim sistemini hedef alan kötücül yazılımların, yapay öğrenme teknikleri ve yaklaşımları kullanılarak tespit edilmesi konuları kapsanmıştır. Bu alandaki bilimsel araştırmalar mevcut literatürden derlenerek ilgili çalışmalarda kullanılan veya kullanılması önerilen tasarım, yöntem ve uygulamalar özet bir biçimde anlatılmıştır., The problem of malware detection, irrespective of the platform of choice, affects nearly all operating systems, including the ones with large user bases such as Windows, Mac, and Linux. A significantly larger market share in the smartphone market compared to even its greatest rival and its open source architecture has made Android operating system the prime target for malware-related threats and cyber-attacks. Therefore, Android became the primary platform for designing and measuring the effectiveness of new approaches and methodologies for malware detection. This literature review focuses on the topic of detection of malware on Android platform by utilizing machine learning techniques and approaches. An extensive collection of the scientific studies on the given topic was collected and the design, the methodology and the real-world applications proposed or implemented by them are described in a short and concise manner.

Bu çalışmada, nitel ve / veya nicel riskleri değerlendirmek ve hesaplamakiçin Bayes ağı ve bulanık çıkarım sistemine dayanan yeni bir bilgi güvenliğirisk değerlendirme yaklaşımı ortaya konmuştur. Önerilen model, bir yazılımşirketinin için test süreçlerini analiz etmek üzere geliştirilmiştir.Tehditler, güvenlik açıkları, riskler ve bunların bağlantılarının tanımlandığıbir Bayes ağı tasarlanmış ve her bir risk faktörü için bileşen olasılıklarıhesaplanmıştır. Bilgi varlıklarının değerleri, riskleri, olasılıkları vegöreceli risk değerleri için bulanık üyelik fonksiyonları ve bulanık kararkuralları tasarlanmış ve oluşturulmuştur. Son aşamada da, risk değerlerininetkileri, bulanık toparlama ve durulaştırma işlemleri ile hesaplanmış vesıralanmıştır. Bu yeni model, kurumlardaki yöneticilerin daha objektif,güvenilir ve esnek bir şekilde bilgi güvenliği risk değerlendirme sonuçlarıelde etmelerine ve kullanmalarına olanak sağlamaktadır., This study proposes a novelinformation security risk assessment approach based on Bayesian network andFuzzy Inference System in order to evaluate and calculate both qualitative and /or quantitative risks. The proposed model is developed to analyse testprocesses for a software services company in order to evaluate the informationsecurity risks. Threats, vulnerabilities, risks, and their relations areconstructed with a Bayesian network and marginal probabilities are calculatedfor each risk factor. Several fuzzy membership functions and fuzzy decisionrules are designed and constructed for assets’ values, risks’ probabilities,and relative risk values. Finally, the impacts of risk values are calculatedafter the aggregation and defuzzification process. It is shown that this new modelenables the business decision makers and managers to obtain more objective,reliable, and flexible information security risk assessment results.

Security risks related to information and information technologies have increased in parallel with technological progress, and knowledge has become the most important asset for individuals and institutions today. During the production, processing, transmission and storage of the knowledge, protecting its confidentiality, integrity and availability has also evolved as an important issue. In this study, the risks faced by the information assets of the organization / institution, the processes affected by the risks and the internal and external parties related to them, the affected assets and locations, the probability of risk, the effect of confidentiality, integrity and accessibility, the value of assets, the probability of creating threats to work effects are defined and a software prototype is designed and developed in order to track and manage the risks and the relevant controls. In order to achieve a proper information security model and construct an information security management system (ISMS) for the organization, the most important factors have been considered. A risk methodology is designed, implemented, documented and risk analysis is conducted on these assets and processes by using these documents. Editor: H. Kemal İlter, Ankara Yıldırım Beyazıt University, Turkey Received: August 19, 2018, Accepted: October 18, 2018, Published: November 10, 2018 Copyright: © 2018 IMISC Durankaya et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Turnaroundtime (TAT) or duration between different stages in medical and healthcareservices is accepted to be one of the most significant performance measuresthat can have a great impact on service quality, change management, costs, andstrategic decisions. Accurate and reliable prediction or estimation of theturnaround times or elicitation of the underlying causes that affect TAT isknown to be a difficult problem. In this study, a heuristic prediction approachis used by designing and implementing a special artificial neural network (ANN)model in order to predict TAT of a specific process in a private hospital. Theprediction performance of our ANN model is comparatively analyzed with somealternative linear and nonlinear numerical prediction algorithms. The resultsshow that ANN surpasses all of the other numerical prediction algorithms andANN might be used by the decision makers as a reliable model to estimate TATwithin acceptable error rates., Hastanelerde ve çeşitli sağlık hizmetlerinde tıbbi işlemler/ aşamalar arasındaki işlem ya da geri dönüş süresi, hizmet kalitesi, değişimyönetimi, maliyetlerin azaltılması ve stratejik kararlar üzerinde de etkisi olanen önemli performans ölçütlerinden biri olarak kabul edilmektedir. Geri dönüşsürelerinin doğru ve güvenilir tahmini ya da bu süreleri etkileyen etmenlerinveya nedenlerin ortaya çıkarılması ise çözümü zor bir problemdir. Bu çalışmada,özel bir hastanedeki çeşitli birimler arasındaki tıbbi iş süreçlerine aitgerçek istatistiksel işlem süresi verileri kullanılarak iş bitirme sürelerininsayısal olarak tahmini için özel bir yapay sinir ağı (YSA) modeli tasarlanmışve kodlanmıştır. YSA modelimizin kestirim performansı, bazı alternatif doğrusal/ doğrusal olmayan sayısal kestirim algoritmaları ile karşılaştırmalı olarakanaliz edilmiştir. YSA'nın tahmin başarısı ve hata değerleri açısından diğertüm sayısal kestirim algoritmalarından daha başarılı olduğu ve YSA'nın, tıbbiiş süreçlerinde iş bitirme sürelerini kabul edilebilir hata oranlarında güvenilebilirşekilde tahmin edebildiği ve karar destek sistemlerinde yöneticiler tarafındanalternatif bir model olarak kullanılabileceği ortaya konmuştur.

Thesis (Master)--Izmir Institute of Technology, Computer Engineering, Izmir, 1999, Includes bibliographical references (leaves: 193-197), Text in English; Abstract: Turkish and English, xii, 204 leaves, The two most well known and recently developed methods in cryptanalysis of DES and DES-like symmetric block ciphers are difTerential and linear cryptanalysis. But these cryptanalytic attacks need to be improved due to the computational performance and storage capacity problems On the other hand, genetic algorithms can be a good solution in cases where the optimum value or near-optimum solutions are sought in complex systems or for non-linear problems. This is a valid situation for the cryptanalysis case where DES and DES-like ciphers are non-linear in structure making dilTerential and linear cryptanalysis a complex system with a very large search landscape and extreme amount of conditional and probabilistic candidates for the key being sought. In this study, a new and promising method wit h bet ter performance is to be developed for differential/linear cryptanalysis of DES and similar symmetric cryptosystems exploiting genetic algorithms' broadened search and optimum finding capacity.

Android mobil işletim sisteminin, rakiplerine kıyasla sahip olduğu oldukça yüksek toplam pazar payının yanında toplamda sayısal olarak çok daha fazla uygulamaya sahip olması dolayısıyla kötücül yazılımlar tarafından en sık hedef alınan mobil platform olduğu bilinmektedir. Son kullanıcının, tipik güvenlik yetersizliğine bağlı olarak, kötücül yazılımın Google Play Store veya herhangi bir resmi olmayan uygulama mağazasında yayımlanmadan önce tespit edilmesi hayati bir öneme sahiptir. Bu tezde, makine öğrenmesi teknikleri kullanarak yeni bir Android kötücül yazılım tespit metodolojisi yanında yeni bir öznitelik seçim metodolojisi ortaya konmuştur. Bu çalışmada sunulan makine öğrenmesi yaklaşımı, Android uygulamalarından (APK dosyaları) statik olarak çıkarılabilen, izinler (permissions), Uygulama Programlama Arayüzü çağrıları (API calls) ve katar (string) özelliklerini kullanmaktadır. Sunulan özellik seçim metodolojisinde literatürdeki mevcut yöntemlerden farklı olarak, belge sıklığı tabanlı (document frequency-based) bir yöntem tasarlanıp uygulanmıştır. Önerilen yöntem, Android kötücül yazılım örnekleri barındıran iki evrensel temel ölçüt veri kümesi ile test edilmiş ve bazı ikili sınıflandırma algoritmaları yanı sıra bazı topluluk (ensemble) yöntemine dayalı algoritmalar da kullanılarak literatürdeki diğer modeller ve yöntemlere göre daha başarılı sayılabilecek yüksek doğrulukta sonuçlar elde edilmiştir. Android is the mobile operating system most frequently targeted by malware in the smartphone market with a significantly higher total market share in comparison to its competitors in addition to a much higher total number of applications. Detection of malware before it is published on the Google Play Store or any unofficial application market is very important owing to the end users' typical security inadequacy. In this Ph.D. thesis, a novel methodology of feature selection is proposed along with an Android malware detection approach that implements the proposed feature selection methodology. The machine learning approach proposed in this thesis makes use of permissions, API calls, and strings as features, which are statically extractable from the Android executables (APK files). In the proposed feature selection approach, a document frequency-based approach was designed and implemented that differs from the existing methods in the literature. The proposed methodology was tested upon two universal benchmark datasets that contain Android malware samples and promising results were obtained by using several binary classification algorithms and some ensemble learning models. 178