Scopus İndeksli Yayınlar Koleksiyonu

Permanent URI for this collectionhttps://hdl.handle.net/20.500.12416/8651

Browse

Filters

2015 - 20205

Settings

Author: search.filters.author.Sever, HayriPublisher: search.filters.publisher.Ieee

Search Results

Now showing 1 - 5 of 5
  • Conference Object
    Citation - WoS: 8
    Citation - Scopus: 14
    Small and Unbalanced Data Set Problem in Classification
    (Ieee, 2019) Sezer, Ebru Akcapinar; Sever, Hayri; Par, Oznur Esra
    Classification of data is difficult in case of small and unbalanced data set and this problem directly affects the classification performance. Small and / or the imbalance dataset has become a major problem in data mining. Classification algorithms are developed based on the assumption that the data sets are balanced and large enough. The most of the algorithms ignore or misclassify examples of the minority class, focus on the majority class. Small and unbalanced data set problem is frequently encountered in medical data mining due to some limitations. Within the scope of the study, the public accessible data set, hepatitis, was divided into small and imblanced data subsets, each of the data subsets were oversampled by distance based data generation methods. The oversampled data sets were classified by using four different machine learning algorithms (Artificial Neural Networks, Support Vector Machines, Naive Bayes and Decision Tree) and the classification scores were compared.
  • Conference Object
    Citation - WoS: 2
    Citation - Scopus: 15
    Securing Blockchain Shards by Using Learning Based Reputation and Verifiable Random Functions
    (Ieee, 2019) Ozsoy, Adnan; Sever, Hayri; Bugday, Ahmet
    In order to meet the increasing demand of the blockchain, it needs to find a solution to the scalability problem. It has been focused on sharding recently to address the scalability problem. In the sharding method, the blockchain is divided into pieces. Instead of a more extensive network, networks with fewer nodes are created. As a result, it becomes more important that each node in the network is reliable. In this study, studies using sharding method have been investigated, and methods for the assigning nodes to shards are proposed. The use of learning-based adaptive methods for this process will contribute to the safe and reliable use of shards. The probability of the shards to deteriorate and influence the whole blockchain will be reduced.
  • Conference Object
    Authorship Modelling Approach for Authorship Verification on the Turkish Texts
    (Ieee, 2018) Akcapinar Sezer, Ebru; Sever, Hayri; Canbay, Pelin; Sezer, Ebru Akcapinar
    Authorship attribution which aims to extract information about an author by analyzing the text of the author is a challenging field that has been studied for years. This study becomes even more difficult when there is limited data on this field. The need for this study carried out under the name of Authorship Verification is increasing day by day with the increase of anonymous authors in the electronic environments. In this study, a model-based solution approach is presented for the authorship verification problem. With the presented approach, it was determined what should be the success interval to be considered in the authorship verification problem.
  • Conference Object
    Citation - Scopus: 2
    Web Service-Based Turkish Automatic Speech Recognition Platform
    (Ieee, 2020) Polat, Huseyin; Sever, Hayri; Oyucu, Saadin
    In response to the similar challenges in building large-scale distributed applications and platforms on the Web, microservice architecture has emerged and gained a lot of popularity in recent years. Therefore, both for the use of microservices and for the provided of the necessary interface for Automatic Speech Recognition (ASR), a web-based platform has been developed. Within firstly the scope of the study, a Turkish ASR system was developed. A web service structure was created to facilitate access to the ASR system. The access of methods and data in the web service structure was provided through Representational State Transfer (REST) web services and service layer. An interface was developed to enable interaction with the web service. The platform was developed using a combination of different technologies such as ASR, web services, microservices, and interface technologies. The developed platform can be used via a standard web browser or an Application Programming Interface (API). In this study, Docker packages were used to improve system performance instead of using different virtual machines on a single server. In the experiments performed, it was shown that the Turkish ASR system had a word error rate of 24.70%. In web service performance tests, it was shown that the platform responded in an average of 9.6 seconds for a 59-second speech recording. The developed user interface was tested in both mobile and desktop web browsers and was shown to function properly. Applications and other services were given access to the platform without the need to use an interface via API support provided by the platform. As a result, a web service-based Turkish ASR platform working seamlessly on the ever-increasing number of mobile devices, the Internet of Things ecosystem, or other access devices was developed.
  • Conference Object
    Citation - WoS: 6
    Citation - Scopus: 7
    Blocked-Dwt Based Vector Image Watermarking
    (Ieee, 2015) Dincer, Kivanc; Sever, Hayri; Elbasi, Ersin; Senol, Ahmet
    Image watermarking is in use for proving ownership for a fairly long time. For most of the study on this area, a pseudo random number sequence PRSN or a binary image logo is embedded as watermark. Nowadays the owner's face or sound is also embedded as biometric watermark. Image is transferred to discrete wavelet transform domain, watermark is embedded to DWT values, then DWT values are retransformed to spatial domain to obtain watermarked image. Embedding a vector image logo as watermark was not tried in previous works. In this work, non-blind robust watermarking is applied using a vector image as watermark. Various attacks are applied to watermarked images and for each of these attacks vector image watermark is obtained equal or almost equal to the original. Embedding vector image as watermark will bring a new discipline for image watermarking and a new development will arise in this perspective.