A gradient boosting classifier for purchase intention prediction of online shoppers

• DOI: 10.1016/j.heliyon.2023.e15163
• Publication Date: 2023-04-03T16:16:37Z
• Authors: Abdullah-All-Tanvir; Iftakhar Ali Khan...; A.K.M. Muzahidul ...; Salekul Islam; Swakkhar Shatabda
• Keywords: Gradient boosting; F1 score; Boosting; Oversampling; Information gain ratio

Early purchase prediction plays a vital role for an e-commerce website. It enables e-shoppers to enlist consumers product suggestions, offer discount and many other interventions. Several work has already been done using session log analyzing customer behavior whether he performs on the or not. In most cases, it is difficult find out make list of customers them when their ends. this paper, we propose customer's intention model where can detect purpose earlier. First, apply feature selection technique select best features. Then extracted features are fed train supervised learning models. classifiers like support vector machine (SVM), random forest (RF), multilayer perceptron (MLP), decision tree (DT), XGBoost have applied along with oversampling method balancing dataset. The experiments were performed standard benchmark Experimental results show that classifier techniques significantly higher area under ROC curve (auROC) score precision-recall (auPR) which 0.937 0.754 respectively. On hand accuracy achieved by Decision improved they 90.65% 90.54% Overall performance gradient boosting compared state-of-the-art methods. addition this, explainable analysis problem was outlined.