Doktora Tezleri

Permanent URI for this collectionhttps://hdl.handle.net/20.500.12469/7776

Browse

Now showing 1 - 3 of 3

Identification of Critical Proteins Associated With Learning Process for Down Syndrome
(Kadir Has Üniversitesi, 2020) Kulan, Handan; Dağ, Tamer
DS protein profilleri laboratuvarda biyokimyasal teknikler uygulayarak gözlemlenmektedir. Fakat, elde edilen protein listesi uzundur ve listedeki her protein DS ile alakalı değildir. Bu yüzden, DS analizi ve tedavisinde, protein ifade miktarları istatiksel metodlar ve makine öğrenmesi teknikleri uygulayarak analiz edilmektedir. Bu tezde, önceki çalışmalara kıyasla, farklı öndeğerlendirme adımları, özellik seçimi ve sınıﬂandırma teknikleri, farklı veri setleri için protein altkümeleri belirlenmesi için uygulanmıştır. Bu protein altkümeleri fareleri daha doğru şekilde ayrıştırır. Spesifik DS özelliklerinin kritik yolaklara etki eden bu altkümelerdeki proteinler tek tek analiz edildiğinde, seçilmiş proteinlerin öğrenme ve hafıza, sinyal yolakları, Alzheimer hastalığı, bağışıklık sistemi ve hücre ölümü gibi önemli süreçlerde rol aldığı gözlemlenmiştir. Bu tezde seçilen protein alt kümelerinden DS un farklı semptomlarını anlamak için yararlanılabilinir ve DS tedavisinde etkili ilaçlar geliştirmek için kullanılabilinir. The protein proﬁles of people with DS are observed by applying biochemical tech niques in laboratory. However, the list of analyzed proteins is long and not all proteins in list are not related to DS. Thus, for the analysis and the treatment of DS, protein expression levels have been analyzed by applying statistical procedures and machine learning techniques. In this thesis, compared to previous works, different preprocessing steps, feature selection and classiﬁcation techniques are applied to deﬁne the subsets of proteins for datasets. These subsets diﬀerentiate mice more accurately. When these subsets which aﬀect the critical pathways of speciﬁc DS aspects are analyzed, it is monitored that selected proteins have vital roles in the processes, such as apoptosis, learning and memory, signaling pathways, immune sys tem and Alzheimers disease (AD). The subsets of proteins selected in this thesis can be applied to interpret the causes of diﬀerent symptoms in DS and can be utilized to foster eﬀective drugs for the cure of DS.
Improving the Quality of Recommender Systems in E-Commerce Platforms
(Kadir Has Üniversitesi, 2021) Işık, Muhittin; Dağ, Hasan
Especially Covid-19 pandemic process, which has taken the world by storm, has shed a clear light on the place of e-commerce, which is already increasing its influence with the globalizing world, in the future world of commerce. This pandemic process has shown that companies that can carry out their business on the internet, regardless of their sector, may survive, and the rest may suffer a great deal. As people choose the way to meet even their daily needs online, it caused the majority of companies to quickly turn to and analyze this field and start to take their place in the e-commerce world as soon as possible. As such, the tricks of e-commerce systems gained great importance. The most prominent of these tricks are product ratings and reviews that completely change the shopping idea of users. The scope of this research consists of experimental studies on how to calculate these product rating systems, which change the profit margin of the companies in the world of e-commerce, more effectively. Our study suggests different methods for the calculation of product rating score to prevent fake accounts, biased or malicious users and companies that make guiding or deceptive interventions for their products and services. That is to say, our study includes alternatives to the primitive calculations used in such systems that can be used in different e-commerce platforms, which can perform various calculations based on reliable users, time factor and reviews of products. In our experimental studies, we have reached various results that can prevent both the deceptive and guiding effect of fake accounts and the primitive and inadequacy of calculation methods in e-commerce systems. In this way, it is aimed that individuals who shop in various e-commerce platforms can reach real information and values and help companies that offer products and services to mirror themselves and create intelligent business ideas.
A Novel Communication Method for Constrained Iot Devices
(Kadir Has Üniversitesi, 2022) Kocatekin, Tuğberk; Tamer Dağ; Cafer Çalışkan
Internet of Things (IoT) is becoming an established part of life by interconnecting billions of devices in diverse areas such as healthcare, smart homes, industries, etc. However, these devices are limited in memory, energy and computational capabilities. Being constrained prevents them from applying complex cryptographic encryption algorithms which leads to lack of security and therefore lack of privacy. As a solution, we propose a novel secret sharing scheme based on underlying protocols of visual cryptography to provide a low-cost and secure communication method for constrained IoT devices. Generally, when a device wants to communicate with an outer party, it does so by itself or by using a mediary such as a central hub or gateway; which leads to single point of failure. As a solution, we propose a method where devices collaborate each other and therefore divide the responsibility into multiple, instead of one. We propose two different models: n-out-of-n and k-out-ofn. In the first model, there is a complete graph where every device is connected to each other. Instead of the original sender, every other device work collaboratively to communicate with the outer party. In the second model, the network is realized as an n-regular graph where a single node has n number of neighbors, which collaborates with each other and here the responsibility is divided into n devices. Results show that this scheme is applicable to constrained devices.

Browse

Browsing Doktora Tezleri by Department "Enstitüler, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Ana Bilim Dalı"