Browsing by Author "Kafkas, Kenan"

Now showing 1 - 4 of 4

Citation Count: 2
Discovering Customer Purchase Patterns in Product Communities: An Empirical Study on Co-Purchase Behavior in an Online Marketplace
(Mdpi, 2021) Kafkas, Kenan; Perdahci, Ziya Nazim; Aydin, Mehmet Nafiz
Marketplace platforms gather and store data on each activity of their users to analyze their customer purchase behavior helping to improve marketing activities such as product placement, cross-selling, or customer retention. Market basket analysis (MBA) has remained a valuable data mining technique for decades for marketers and researchers. It discovers the relationship between two products that are frequently purchased together using association rules. One of the issues with this method is its strict focus on binary relationships, which prevents it from examining the product relationships from a broader perspective. The researchers presented several methods to address this issue by building a network of products (co-purchase networks) and analyzing them with network analysis techniques for purposes such as product recommendation and customer segmentation. This research aims at segmenting products based on customers' purchase patterns. We discover the patterns using the Stochastic Block Modeling (SBM) community detection technique. This statistically principled method groups the products into communities based on their connection patterns. Examining the discovered communities, we segment the products and label them according to their roles in the network by calculating the network characteristics. The SBM results showed that the network exhibits a community structure having a total of 309 product communities, 17 of which have high betweenness values indicating that the member products play a bridge role in the network. Additionally, the algorithm discovers communities enclosing products with high eigenvector centralities signaling that they are a focal point in the network topology. In terms of business implications, segmenting products according to their role in the system helps managers with their marketing efforts for cross-selling, product placement, and product recommendation.
Citation Count: 0
Ground Truth in Network Communities and Metadata-Aware Community Detection: a Case of School Friendship Network
(2021) Perdahçı, Ziya Nazım; Aydın, Mehmet Nafiz; Kafkas, Kenan
Real-world networks are everywhere and can represent biological, technological, and social interactions. They constitute complicated structures in terms of type of things and their relations. Understanding the network requires better examination of the network structure that can be achieved at various scales including macro, meso, and micro. This research is concerned with meso scale for a student best friendship network where sub-structures in which groups of entities (students) take different functions. In this study we address the following research questions: To what extent would NeoSBM as a stochastic process underlie best friendship interaction and in turn ground truth interactions (i.e. reported best friendship)? Do metadata such as gender or class contribute to this understanding? How can one support school managers from a meta-data aware community detection perspective? Our findings suggest that metadata aware community detection can be an effective method in supporting decision-making for class formation and group formation for in and out school activities.
A Network Science Approach To Correlations Between Course Achievement and Community Structure in School Friendship Networks
(Kadir Has Üniversitesi, 2017) Kafkas, Kenan; Aydın, Mehmet Nafiz
in this research we examine a secondary school social network. We apply social network analysis (SNA) techniques on the close friendship structure of the students. Our aim is to answer the following questions regarding the network. The first question is what are the mixing values with respect to test achievement scores gender and class. The second question is how are the communities in the network structured. Such findings can be significant assets in understanding and improving the learning environment. They may be used to help teachers and school managers in deciding more workable and efficient student matchings. For this study we conducted a survey to a group of 10th grade students and gathered the necessary information to construct the social network around the school. Our findings show that the friendship in overall network is neither assortative nor disassortative with respect to academic success in other words the two attributes are not correlated. On the other hand gender and class mixing measures are significantly high which not surprisingly suggests that the students prefer to bond with their classmates and also with the same gender friends. Finally after examining the communities within each classroom we observe similar community structures. in the light of these findings we propose a method for composing the classrooms to construct an efficient and successful learning environment.
Product and Customer Segmentation by Purchase Behavior in E-Commerce Platforms Using Stochastic Block Model
(Kadir Has Üniversitesi, 2022) Kafkas, Kenan; Aydin, Mehmet; Perdahçı, Ziya Nazım
To attract and maintain lucrative clientele, commercial internet platforms compete with a multitude of competitors by providing appropriate goods and services, em ploying a range of marketing methods to get a competitive edge utilizing their digital trace data. Techniques include a variety of marketing tactics, many of which are based on updated versions of conventional marketing strategies. Working out what consumers want and how to meet their needs is an ongoing task on these platforms. The literature is constantly being enhanced by new theoretical and practical applica tions. Customer purchase behavior leaves digital trace data in online platforms such as clickstream, transaction, or product review forms. This thesis proposes a model that presents a novel network approach to customer behavior analytics on online transaction data to perform product and customer segmentation. We seek answers to the following research questions: Can we understand the customer behavior and preferences through network analysis? If there are several purchase behavior types, what are the underlying patterns? Are there certain special products that play a special role in the network? To support decision-makers in their endeavor to improve marketing activities such as targeted advertising, increasing brand loyalty, attract ing desired customers, and signaling more effective marketing messages. We utilize the Stochastic Block Model (SBM), which is a statistically principled community detection method on co-purchase networks to discover latent product communities, and we produce two different segmentation methods based on those communities. The outcome is a product and a customer segmentation which extends traditional data mining methods. We combine product based segmentation with Market Bas ket Analysis and customers segmentation with the RFM models. We implement our model on two empirical data sets. Lastly, we provide an executive summary for both examples.