izpis_h1_title_alt

Churn prediction with machine learning methods
ID ANGELOVSKA, BLAGICA (Author), ID Marolt, Matija (Mentor) More about this mentor... This link opens in a new window

.pdfPDF - Presentation file, Download (955,57 KB)
MD5: B923B1326B8141F7DA1A8E8247A52DCF

Abstract
Churn rate affects all businesses, from the smallest to the largest companies. It has a big role in the CRM strategy, where it is usually used for analyzing the return on marketing investments. In the gambling industry, it is important to analyze and compare the detected churners, so the marketeers can cleverly use the marketing efforts. Predicting churn rate will not only minimize the outflow of customers, but will also increase their time spent at the casino. In order to be able to determine it, it is essential to understand each individual customer better. In this thesis, we analyze the gathered data on casino players, define a churn definition and use supervised learning methods for predicting customer churn. We focus on gradient boosted trees, more precisely XGBoost (eXtreme Gradient Boosting trees), and explore how their structure and parameters affect the prediction. Through analysis, we proved that by using demographic and visiting information of customers, it is possible to successfully predict the potential churn customers.

Language:English
Keywords:machine learning, classification model, customer attrition, marketing
Work type:Bachelor thesis/paper
Organization:FRI - Faculty of Computer and Information Science
Year:2019
PID:20.500.12556/RUL-109930 This link opens in a new window
Publication date in RUL:10.09.2019
Views:1202
Downloads:334
Metadata:XML RDF-CHPDL DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:Slovenian
Title:Napovedovanje odhodov strank z metodami strojnega učenja
Abstract:
Stopnja odhodov strank vpliva na vsa podjetja. Ima veliko vlogo v strategiji odnosov s strankami, saj je pomembno analizirati stopnjo odhodov in vzroke za odhode, da lahko tržniki pravilno usmerijo trženjska prizadevanja. V igralniški industriji napovedovanje odhodov strank ne bo samo zmanjšalo njihovega odliva, temveč bo lahko tudi povečalo čas, preživet v igralnici. Da lahko odhode napovemo vnaprej, je nujno, da vsako posamezno stranko bolje razumemo. V diplomski nalogi obravnavamo podatke o strankah v igralnicah, definiramo ciljno spremenljivko, ki določi kdaj se je zgodil odhod stranke, in uporabimo nadzorovane metode strojnega učenja za napovedovanje ciljne spremenljivke. Poudarek je na odločitvenih drevesih, natančneje algoritmu XGBoost, kjer smo preučili kako drevesna struktura in parametri algoritma vplivajo na kvaliteto napovedi. Z analizo ugotavljamo, da je mogoče z uporabo demografskih podatkov in informacij o obiskih dokaj uspešno napovedati odhode strank.

Keywords:strojno učenje, klasifikacijski model, odhajanje strank, trženje

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back