<![CDATA[Customer retention and monetization have since been the pillar of many successful firms and businesses as keeping an old customer is far more economical than gaining a new one – which, in turn, reduce customer churn rate. Previous studies have focused on the use of single heuristics as well as provisioned no retention strategy. To curb this, our study posits the use of the recen-cy-frequency-monetization framework as strategy for customer retention and monetization impacts. With dataset retrieved from Kaggle, and partitioned into train and test dataset/folds to ease model construction and training. Study adopt a tree-based Random Forest ensemble with synthetic minority oversampling technique edited nearest neighbor (SMOTEEN). Various benchmark models were trained to asssess how well each performs against our proposed ensemble. The application was tested using an application programming interface Flask and integrated using streamlit into a device. Our RF-ensemble resulted in a 0.9902 accuracy prior to applying SMOTEENN; while, LR, KNN, Naïve Bayes and SVM yielded an accuracy of 0.9219, 0.9435, 0.9508 and 0.9008 respectively. With SMOTEENN applied, our ensemble had an accuracy of 0.9919; while LR, KNN, Naïve Bayes, and SVM yielded an accuracy of 0.9805, 0.921, 0.9125, and 0.8145 respectively. RF has shown it can be implemented with SMOTEENN to yield enhanced prediction for customer churn prediction using Python]]>
Copyrights © 2024
