SEARCH RESULT

Co-authorship networks provide a broad view to the connectivity properties of scholars, together with patterns of knowledge diffusion in scientific society. Network science provides a substantial framework for discovering the dynamics of these interactions those are defined by co-authoring a paper together. We constructed a complex network consisting of co-authorship links between authors, using the data retrieved from Web of Science Core Collection. Date retrieved is limited to 67248 publications addressed from Turkey in engineering field, including the timespan between 1975 and 2018. Analysis performed through this massive dataset resulted a complex network of 78883 nodes (authors) and 194232 edges (co-authorship links). Authors exhibit an average degree (neighbor) of 4.925, which increases to 6.687 in weighted analysis. Network exhibits an invincible clustering coefficient of ~0.8, while the average path length is close to 18. Together with the power-law consistent degree distribution that labels the network as scale-free, we also presentedtop “most central” authors of this network with respect to betweenness, closeness and eigenvector centrality measures, each defining the “importance” of an author in different aspects.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

İlker TÜRKER Rafet Durgut Oğuz Findik

Recognition and interpretation of human activities are very interesting and hot topics that are frequently studied in the field of computer vision. Especially with the advent and development of the Microsoft Kinect depth sensors, the expansion of the study fieldhas gained momentum in the positive direction. Thanks to RGBD cameras, which also provide depth information in addition to the RGB image, researchers benefit from many advantages in terms of privacy, accuracy and precision. In this study, automatic segmentation of repeated 3D human activity is proposed. A public dataset containing the repeated action sequences are recorded using the RGBD camera. The action sequence in this dataset includes similar and different action information. In order to identify and label each action in sequence, it is necessary to perform the segmentation process. To be able to perform a successful segmentation process, the data must be preprocessed to remove noise. For this purpose, a total variation based noise removal method is used. Human action recognition and detailed error analysis can be performed through the segments derived from the output of this work.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Rafet Durgut C. OZCAN Oğuz Findik

In parallel with the development of the technology, the storage of the data more easily and quickly and the faster processing on the stored data can make an important contribution to the creation and analysis of networks of coexistence. Complex networking plays an important role in analyzing and revealing common characteristics and structures of connected clusters depending on various characteristics. In this study, a network of association between male tennis player in the Australian Open, the French Open, the US Open and the Wimbledon tennis tournaments, known as four major international tennis tournaments between 2000 and 2017, has been established.While each tennis player is defined as a node in the network of associations created, the competitions of the tennis players with each other are defined as the links connecting these nodes. The universal principles of complex networks such as scale-free, small world, clustering have been examined. Furthermore, through Gephi software, the structural characteristics of networks are visualized by using the data obtained from this association network. As a result of the study, it was seen that the networks among the tennis players struggling in the related tournaments were carrying real world network characteristics and that the data obtained from these networks can be used for network analysis.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Oğuz Findik Emrah Özkaynak

Social media platforms such as Twitter have grown at a tremendous pace in recent years and have become an important source of data providing information countless field. This situation was of interest to researchers and many studies on machine learning and natural language processing were conducted on social media data. However, the language used in social media contains a very high amount of noisy data than the formal writing language. In this article, we present a study on diacritic restoration which is one of the important difficulties of social media text normalization in order to reduce the noise problem. Diacritic is a set of marks used to change the sound values of letters and is used on many languages besides Turkish. We suggest a 3-step model for this study to overcome the top of the diacritic restoration problem. In the first stage, a candidate word producer produces possible word forms, in the second stage the language validator chooses the correct word forms and at the final word2vec is used to create vector representations of the words and make the most appropriate word choice by using cosine similarities. The proposed method was tested on both synthetic and real data sets, and we achieved a relative error reduction of 37.8% in our data sets compared to the previous study with an average of 94.5% performance.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Zeynep Ozer İlyas özer Oğuz Findik

In this paper, one-shot-learning gesture recognition methods are reviewed and an approach of hand gesture recognition using one-shot-learning is proposed. This approach aims to recognize new categories of gestures from a single video clip of each gesture. The gestures are generally related to a particular task, for instance, hand signals used by divers, finger codes to represent numerals, etc. In this study, both RGB and depth images are utilized for a given dataset. A rich dataset, namely the ChaLearn Gesture Dataset (CGD2011), are employed. The dataset is divided into 20 different files which include 940 videos in total. Although training the system with only one example is difficult, depth and RGB images provide many new possibilities. We used the standard deviation of the depth images of a gesture and motion history image (MHI) method. Also, two dimensional fast fourier transform (2D FFT) is used to reduce the effect of camera shift. It is seen that FFT has no distinct effect on the image quality. Then, we compare image templates based on the correlation coefficients and Levenshtein, Mahalanobis, Frobenius distance measures. The Levenshtein distance measure is more suitable to match image templates compared to other distance measures. It is observed that MHI method gives better hand gesture recognition accuracy about one-shot-learning.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Esma Şeker Oğuz Findik

Recently, through development of several 3d vision systems, widely used in various applications, medical and biometric fields. Microsoft kinect sensor have been most of used camera among 3d vision systems. Microsoft kinect sensor can obtain depth images of a scene and 3d coordinates of human joints. Thus, anthropometric features can extractable easily. Anthropometric feature and 3d joint coordinate raw datas which captured from kinect sensor is unstable. The strongest reason for this, datas vary by distance between joints of individual and location of kinect sensor. Consequently, usage of this datas without kinect calibration and data optimization does not result in sufficient and healthy. In this study, proposed a novel method to calibrating kinect sensor and optimizing skeleton features. Results indicate that the proposed method is quite effective and worthy of further study in more general scenarios.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Mahmut Selman Gökmen Mehmet Akbaba Oğuz Findik

Online marketing platforms have turned into large volumes of information and opinion for customers with the transition to Web 2.0. Customers refer to these resources in order to obtain information before they purchase a product and to reach the potential views of others about possible experiences. Businesses also need customer feedback to improve the services they provide and to explore which reviews are more valuable product specifications. In this study, Sentiment Analysis (SA) was performed with 2-pole (positive-negative) classification about hotel businesses on an opinion dataset created by users. Deep Learning based Recurrent Neural Network (RNN) architecture was used in these analyzes. With the results of the RNN architecture, the results of the classification based on score conditional and editorial interpretation were compared and it was observed that the performance of classification with RNN was successful.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Kürşat Mustafa KARAOĞLAN Volkan Temizkan Oğuz Findik

The research study predicts surgical operations based on patient characteristics using different types of Learning Vector Quantization algorithms. The primary goal is to identify whether a patient requires surgery or not and classify the type of surgery needed. The paper utilizes a disease dataset containing many patient attributes, including disease-specific factors and medical history to train and evaluate the models. Also, tested types of LVQ algorithms including LVQ, RSLVQ, Soft LVQ (SLVQ), Generalized LVQ (GLVQ), Fuzzy LVQ, and LVQ3. Results show that GLVQ achieved the highest performance with an accuracy of 98.42%, precision of 0.99, recall of 0.97, and F1- score of 0.98. The discovery shows that advanced GLVQ can be very useful in healthcare for making predictions. This model can help doctors make better decisions by accurately predicting whether a patient needs surgery.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Ali Asghar Oğuz Findik Emrah Özkaynak

Bu çalışmada; Yapay Sinir Ağı - Çok Katmanlı Algılayıcı (YSA-ÇKA) ve Çoklu Regresyon Analizi (ÇRA) yöntemleri kullanılarak, Türkiye’de demir çelik endüstrisine ev sahipliği yapan başlıca iller arasında yer alan Karabük kentsel alanına ilişkin, günlük Partikül Madde (PM10) kirletici gösterge seviyesinin tahminine yönelik bir yaklaşım sunulmuştur. Söz konusu yaklaşımda, veri seti olarak 2005 ile 2015 yıllarına ilişkin günlük olarak ölçülmüş, meteorolojik gözlem verileri ve kirletici değerler kullanılmıştır. Gerçekleştirilen deneysel çalışmalar sonucunda; hava kirletici değerlerin tahmininin mümkün olduğu sonucuna ulaşılmıştır. Ayrıca uygulanan YSA-ÇKA ve ÇRA yöntemleriyle elde edilen sonuçların performansları da değerlendirilmiştir. Deneysel çalışmalar sonucunda, YSA-ÇKA yönteminin, ÇRA’ya göre daha iyi bir performans gösterdiği sonucuna ulaşılmıştır.

Akademik Bilişim
AB

Kürşat Mustafa KARAOĞLAN Ü. ATİLA Yusuf Kurtgoz Oğuz Findik