Please use this identifier to cite or link to this item: https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4639
Full metadata record
DC FieldValueLanguage
dc.contributor.authorSandamali, G.A.H.-
dc.date.accessioned2022-08-26T04:43:24Z-
dc.date.available2022-08-26T04:43:24Z-
dc.date.issued2022-08-26-
dc.identifier.urihttps://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/4639-
dc.description.abstractAlmost all the business entities thrive their businesses by attracting the customers largely to their businesses day by day. For that, they introduce attractive service campaigns, marketing techniques, and advertising as well. But after all, the business giants found that the churning of the customers from their products or the services can affect largely the profits of their businesses. Hence their existence in the business world depends on the number of customers who retain their services not merely the number of consumers at the beginning. So, out of all the business domains, this research focuses on the credit card domain which largely affects by the churning of their customers. With the literature, the research identified the number of single machine learning models that were mostly used related to retention prediction. Afterward, the single machine learning models such as Logistic Regression, Random Forest, MLP, k-Nearest Neighbor, Naïve Bayes, Decision Tree, Ada Boost, XGBoost, and LightGBM are used and the performance is evaluated using the metrics such as Accuracy, Area under Curve and Mathews Correlation. Then with the comparison of their performance identification of the weak learners and the meta-model is achieved. Then an ensemble model is created by stacking the weak learners with the meta-model to achieve the increased performance rather than that of the single model. The built ensemble model guarantees the accuracy of 0.9645, the AUC value of 0.9455 and along with the other performance metrics proving that the ensemble machine learning model is the best solution from the rest of the single models that were being used. Through this research, it is identified that Total transaction count, the ratio of Total Transaction count over quarter 4 vs quarter 1, Total Revolving Balance, Average Utilization Ratio, and Total Transaction Amount are the features that positively correlated with retention of the credit card usersen_US
dc.language.isoen_USen_US
dc.subjectdata miningen_US
dc.subjectmachine learningen_US
dc.subjectcredit card consumer attritionen_US
dc.subjectensemble modelen_US
dc.subjectstackingen_US
dc.titleRetention Prediction of Credit Card Users Using Data Mining and Machine Learning Techniquesen_US
dc.typeThesisen_US
Appears in Collections:2021

Files in This Item:
File Description SizeFormat 
2018 MCS 079.pdf2.47 MBAdobe PDFView/Open


Items in UCSC Digital Library are protected by copyright, with all rights reserved, unless otherwise indicated.