A machine learning based integrated framework for e-commerce management through customized customer retention strategy

Abstract

Ine-commerce,itismorecost-effectivetoretainexistingcustomersthantoattractnewones. Therefore, churn control is crucial in managing attrition. This study aims to ad-dressagapintheliteraturebycombiningchurnprediction,customersegmentation,andrecommendationstoprovideacomprehensiveplatformforcustomizedcustomerretentionstrategy.Theframeworkconsistsofthesethreecomponents.ExperimentalanalysisforcustomerchurnpredictionshowsthatCatBoostperformsthebestinthedatasetwith100%accuracyand100%F1-score.Afterselectingthebestclassifier,recursivefeatureelimination (RFE) is applied to find the rank of features for the next step. This paperfills a research gap and contributes to the existing literature in developing a customerretentionmethod. After implementing churn prediction, the cluster tendency test is performed, and theHopkins score is 0.09322634776929459. This score indicates that the dataset is appropriateforclustering.TheCalinski-HarabaszindexofK-meansishigherthanHierarchical Clustering.Comparing k-means, hierarchical clustering, and DBSCAN, k-meansgives the best performance.The resulting cluster profiles are used for personalizedrecommendations. Collaborative filtering coverage is larger than popularity-based recommendations.The root mean square errors (RMSE), mean absolute errors (MAE),and mean square errors (RMSE) are used for comparing criteria. The SVD model hasRMSE 1.26, MAE 0.99, and MSE 1.69. The test time of SVD is 0.00143, which is thelowest.Precisionis89%andrecallis99.80%whichmeansthat89%oftherecommendations apply to users. Our proposed study shows that with this integrated framework,wehaveachievedsignificantimprovement.Ourstudyhasgiventhebestresultineveryphase of the framework. So, integrating product recommendation with customer churnpredictionsandcustomersegmentationwillenablepersonalizedexperiences,proactivecustomer engagement, maximization of customer lifetime value, data-driven decision-making, and a competitive edge in the market in the context of our country.It willalsoprovideastrongandefficientcustomerretentionstrategytomakee-commercefeelsuccessfulinourcountry.Finally,thisresearchinvestigatesthekeysustainabilityindicatorsforintegratingMLine-commercecustomizedcustomerretentionandconductsasystematicassessmenttoprioritizetheindicatorsbasedontheperspectivesofpertinentexperts in the context of Bangladeshi e-business to assess the long-term sustainabilityoftheproposedsystem.