Research on User Behavior Prediction of E-Commerce Platform Based on Transformer-LSTM Algorithm

本文探讨了在电子商务领域，利用Transformer-LSTM混合神经网络算法对用户行为进行预测的方法。随着网络购物的普及，用户浏览、收藏、加入购物车等在电商平台产生的大量行为数据成为预测其购买行为的关键信息。在处理此类序列性和依附关系复杂的数据时，传统的机器学习方法有一定的局限性。因此，本文提出了一种结合Transformer和LSTM的混合神经网络模型，以提高预测的准确性和效率。Transformer模型擅长在序列数据中捕捉长时间的依赖关系和并行计算，而LSTM则可以针对梯度消失和爆炸等问题，对长序列数据进行有效处理。本文的模型首先通过Transformer对用户行为数据进行编码，提取关键特征，然后利用LSTM对提取的特征序列进行建模，最终实现对用户购买行为的预测。实验结果显示，该模型比传统的LSTM模型在精确度、精确率、召回率、F1-score等指标上都要好，在电商用户行为预测上显示出潜力和优势。电商平台通过这种改良的模式，能够将商品更精准地推荐给用户，提升用户的购物体验，同时也提高了平台的经济效率。This paper discusses the method of predicting user behavior by using the Transformer-LSTM hybrid neural network algorithm in the field of e-commerce. With the popularity of online shopping, the large amount of behavioral data generated by users on e-commerce platforms (such as browsing, collecting, adding to shopping carts, etc.) has become the key information for predicting their purchasing behaviors. Traditional machine learning methods have certain limitations when dealing with such data that have complex sequentiality and dependencies. Therefore, this paper proposes a hybrid neural network model combining Transformer and LSTM to improve the accuracy and efficiency of prediction. The Transformer model is good at capturing long-term dependencies and parallel computing in sequence data, while LSTM can effectively handle long sequence data and solve the problems of vanishing and exploding gradients. The model in this paper first encodes the user behavior data through Transformer to extract key features, and then uses LSTM to model the extracted feature sequence, ultimately achieving the prediction of user purchasing behavior. The experimental results show that this model outperforms the traditional LSTM model in terms of indicators such as accuracy rate, precision rate, recall rate and F1-score, demonstrating its potential and advantages in the prediction of e-commerce user behaviors. Through this improved model, e-commerce platforms can recommend products to users more accurately, enhancing the shopping experience of users and the economic benefits of the platform.