Customer Behaviour Analysis and Personalized Marketing Using Bidirectional Long Short-Term Memory with Extreme Gradient Boosting

Le résumé fourni par la source

In recent years, e-commerce business evolved rapidly due to rise in smart shopping platforms and Artificial Intelligence (AI) advertisements. Predicting the customers purchase intent is important in this era for recommending best products and increasing purchase conversions. Even though, previous researchers have suggested various advanced model, but still there are challenges like inability to handle complex customer interactions and identify meaningful patterns in sequential data. Therefore, this research proposes Bidirectional Long Short-Term Memory with eXtreme Gradient Boosting (BiLSTM-XGBoost) for accurately predicting customer purchase intent. Initially, data is collected from publicly available e-commerce customers sales-record dataset, which consists customers past purchasing behaviour data. This collected data is pre-processed by using Min-Max normalization for ensuring all numerical features are transformed within a standardized range, and using mahalanobis distance for outlier detection. After that, Recursive Feature Elimination (RFE) with Support Vector Machine (SVM) is employed for selecting relevant features from pre-processed data. Finally, BiLSTMXGBoost is employed, in which BiLSTM obtains sequential patterns from customer behaviour by capturing both past and future dependencies while XGBoost classifies data for accurate purchase intent predictions. The proposed BiLSTM-XGBoost achieved better results in terms of accuracy (98.15%), precision (98.50%), Recall (97.81%) and F1-Score (97.11%) respectively when compared to existing LSTM.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.