Πρόβλεψη απώλειας πελατών για εμπορικές επωνυμίες καταναλωτικών αγαθών με τη χρήση δεδομένων ηλεκτρονικών συναλλαγών

Παπαδάκη, Μαρία - Μαλεβή

Master Thesis

Author

Παπαδάκη, Μαρία - Μαλεβή

Date

2024-02

Abstract

This master's thesis delves into the realm of Customer Churn Prediction, focusing on the novel approach of forecasting potential customer churn at the level of consumer goods brands in the context of electronic commerce. The central aim is to identify customers likely to disengage from a specific brand, offering insights valuable for product manufacturers to shape future strategies and promotional actions in collaboration with retailers. Drawing upon real transactional data from a prominent international online pharmacy, the study refrains from including demographic, geographic, or personal customer information and uses solely transactional data instead. It leverages SQL for data acquisition and Python for dataset construction, machine learning model training, and result visualization. The investigation reveals that predicting customer churn at a brand level is possible, opening new avenues for research, departing from conventional approaches that primarily address businesses or electronic stores. Several machine learning algorithms and feature selection techniques are tested, leading to a final predictive model with an accuracy nearing 70%. In conclusion, this research not only demonstrates the efficacy of predicting customer churn at the brand level but also proposes practical applications, such as integration into the eRAM platform. Furthermore, it outlines future pathways for enhancing predictive accuracy and broadening the scope of the model's utility.

Postgraduate Studies Programme

Πληροφοριακά Συστήματα και Υπηρεσίες

Department

Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών Συστημάτων

Number of pages

Language

Greek

URI

https://dione.lib.unipi.gr/xmlui/handle/unipi/16385
http://dx.doi.org/10.26267/unipi_dione/3807

Collections

Τμήμα Ψηφιακών Συστημάτων

Show full item record

Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα