AaPiDL: an ensemble deep learning-based predictive framework for analyzing customer behaviour and enhancing sales in e-commerce systems

Customer behaviour analysis is essential for optimizing the performance and efficiency of recommender systems. With rapid advancements in deep learning (DL) algorithms, the potential for understanding and predicting customer preferences has significantly improved. Therefore, the given paper proposes an ensemble framework for analysing customer behaviour in order to enhance satisfaction experience and personalized recommendations. The proposed framework consists of dual modules namely purchasing intention (PI) and abandonment analysis (AA) which utilizes amalgamation of three deep learning algorithms namely convolutional neural network (CNN), generative adversarial network (GAN) and long short-term memory recurrent neural network (LSTM-RNN). This ensemble prediction approach provides valuable insights into customer's behaviour and interaction with E-commerce websites. Lastly, the performance of the proposed framework is validated with existing recent studies based on evaluation metrics such as precision, recall, accuracy and F1-score. The framework is intended to provide improved accuracy and generalization capabilities thus enabling better recommendations tailored to individual customer preferences.