Scholarlink

<p>In recent years, the rapid growth of the Internet economy has driven a sharp increase in global cross-border e-commerce sales, presenting vast market potential. However, compared with domestic e-commerce, cross-border platforms face greater challenges in logistics, warehousing, and supply management, while product sales abroad are more volatile due to user preferences, cultural differences, and environmental factors. To address these issues, this paper develops a Gradient Boosting Decision Tree (GBDT) -based sales prediction model by selecting key influencing factors from historical sales data, product information, and customer reviews. For comparison, Backpropagation (BP) neural network and Autoregressive Integrated Moving Average (ARIMA) models are also introduced. Experimental results show that the GBDT model achieves a significantly lower average relative error than the other two models and performs better than the actual prediction performance of Company M, demonstrating that the proposed method improves forecasting accuracy and holds practical application value.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning–driven sales forecasting for cross-border e-commerce platforms

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月5日

查看全文
  • Qiying Lei

摘要

In recent years, the rapid growth of the Internet economy has driven a sharp increase in global cross-border e-commerce sales, presenting vast market potential. However, compared with domestic e-commerce, cross-border platforms face greater challenges in logistics, warehousing, and supply management, while product sales abroad are more volatile due to user preferences, cultural differences, and environmental factors. To address these issues, this paper develops a Gradient Boosting Decision Tree (GBDT) -based sales prediction model by selecting key influencing factors from historical sales data, product information, and customer reviews. For comparison, Backpropagation (BP) neural network and Autoregressive Integrated Moving Average (ARIMA) models are also introduced. Experimental results show that the GBDT model achieves a significantly lower average relative error than the other two models and performs better than the actual prediction performance of Company M, demonstrating that the proposed method improves forecasting accuracy and holds practical application value.