Özet:

Purpose- The study investigates the effect of data-based marketing, which is of great importance for today's businesses, on the creation of customer segments and on the development of marketing strategies for those segments. The big data study, which consists of real customer data analyzed in this direction, aims to identify consumer behaviors that are not similar enough to develop different strategies and to determine how the strategy development processes for customer segments can be done analytically. Methodology- It was studied according to the 2018 order data, which is a data set of 24 million lines in total from an international pizza brand operating in the home delivery service field across the Turkey. In the study, K Means, Gaussian Mixture and DBSCAN algorithms are used for customer segmentation. The clustering and multiple regression analyzes were applied with the Phyton program. Findings- In this study, in which the most used clustering algorithms in the literature were tested, due to the fact that the DBSCAN algorithm is not suitable for the data set used in the application, 91% of all data is assigned to a cluster in the cluster analysis and the remaining data are classified outlier. Accordingly, it has been observed that algorithms such as K Means or Gaussian Mixture give better results in studies where there is no demographic data and behavioral characteristics form the main mass of the data. In addition, clusters exhibiting similar behavior were identified in multiple regression analyzes, in which the forward-looking behaviors of the formed clusters were analyzed, and valuable clusters in the sub-clusters were discovered. Conclusion- While this study covers the steps of big data and data mining, it also covers all end to end processes with multiple regression analyzes to create customer clusters and determine the future behavior models of the determined clusters. In this direction, it offers practitioners an exemplary model and strategy determination methodology.

Anahtar Kelimeler: