Browsing by Author "Kariniauskaite, Dzordana"

Now showing 1 - 2 of 2

Modularity analysis of a bipartite network for an e-commerce shop
(Kadir Has Üniversitesi, 2016) Aydın, Mehmet Nafiz; Aydın, Mehmet Nafiz
Many real-world systems which are of interest to both researchers and practitioners can be modeled as networks – sets of nodes, representing objects, and links between them, representing the interactions among these objects. One of the most important categories of complex networks in naturally real-world systems is bipartite networks (opposite to general unipartite networks), where nodes can be divided into two disjoint sets such that no two nodes of the same type are connected; there are no links connecting nodes of the same type. The identification of communities in networks is crucial for understanding its underlying structure and behavior. In this study, the bipartite network of Internet shop web platform, where buyers and products represent nodes and purchases made represent links, is analyzed. The analysis is based on the modularity function by means of an open source network analysis and visualization tool Gephi. The twenty biggest modules, including hubs, of the giant component are analyzed in depth. The results of the analysis of category types of product hubs could be used for creating new type of product categories in the e-shop, where the product categories are formed according the most popular product types between communities, leaving behind the traditional marketing methods when the product groups are created considering the characteristics and similarities of the products or the most bought products in the e-shop.
Citation Count: 1
Validity Issues of Digital Trace Data for Platform as a Service: A Network Science Perspective
(Springer international Publishing Ag, 2018) Aydın, Mehmet Nafiz; Kariniauskaite, Dzordana; Perdahci, N. Ziya
Data validity becomes a prominent research area in the context of data science driven research in the past years. In this study, we consider an application development on a cloud computing platform as a promising research area to examine digital trace data belonging to records of development activity undertaken. Trace data display such characteristics as found data that is not especially produced for research, event-based, and longitudinal, i.e., occurring over a period of time. Having these characteristics underlies many validity issues. We employ two application development trace data to articulate validity issues along with an iterative 4-phase research cycle. We demonstrate that when working with digital trace data, data validity issues must be addressed; otherwise it can lead to awry results of the research.