Zhao Cheng, Xun Yi
School of Art & Design, Guangdong University of Technology, Guangzhou, 510000, China.
Sci Rep. 2025 Jul 1;15(1):21724. doi: 10.1038/s41598-025-06689-0.
This work aims to accurately understand user satisfaction in online shopping, reflecting user preferences and promoting the development of online shopping. This work explores a behavioral prediction method for online shopping users using a Recurrent Neural Network (RNN) model. Traditional RNN faces challenges in training on long sequences and is susceptible to the vanishing gradient problem. To address this problem, the proposed Gated Recurrent Unit (GRU) introduces a gating mechanism to capture long-term dependencies in sequential data. Building on this, a Dynamic Weighted-GRU (DW-GRU) model is proposed, incorporating a dynamic weighting mechanism based on GRU to adapt to the dynamic changes in online shopping satisfaction. This improvement allows the model to effectively learn and remember long-term dependencies in sequential data while alleviating the vanishing gradient problem. Experimental evaluations of the prediction model are conducted on an Amazon shopping dataset. Comparative analysis reveals that the DW-GRU model outperforms the standard RNN model, showing lower errors with accuracy, precision, and recall values of 0.871, 0.667, and 0.667, respectively. The research findings provide valuable guidance for operators and data analysts of online shopping platforms, furnishing feasible technical support for enhancing user satisfaction and delivering precise product recommendations.
这项工作旨在准确了解在线购物中的用户满意度，反映用户偏好并促进在线购物的发展。这项工作探索了一种使用循环神经网络（RNN）模型对在线购物用户进行行为预测的方法。传统的RNN在长序列训练中面临挑战，并且容易受到梯度消失问题的影响。为了解决这个问题，提出的门控循环单元（GRU）引入了一种门控机制来捕捉序列数据中的长期依赖关系。在此基础上，提出了一种动态加权GRU（DW-GRU）模型，该模型基于GRU纳入了一种动态加权机制，以适应在线购物满意度的动态变化。这种改进使模型能够有效地学习和记住序列数据中的长期依赖关系，同时缓解梯度消失问题。在亚马逊购物数据集上对预测模型进行了实验评估。比较分析表明，DW-GRU模型优于标准RNN模型，其准确率、精确率和召回率分别为0.871、0.667和0.667，误差更低。研究结果为在线购物平台的运营商和数据分析师提供了有价值的指导，为提高用户满意度和提供精准的产品推荐提供了可行的技术支持。
