Scholarlink

<p>This study develops an intelligent e-commerce talent demand prediction system built on a hybrid Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) architecture, designed to support vocational college program optimization under China’s Double First-Class Initiative. The hybrid model pairs LSTM networks for temporal sequence modeling with CNN components for spatial feature extraction, processing multi-dimensional data drawn from job platforms, industry reports, and employment statistics. Experimental validation confirms that the hybrid approach outperforms traditional methods, with a 32.4% reduction in MAE, a 28.7% improvement in RMSE, and an R<sup>2</sup> coefficient of 0.891—compared to standalone LSTM (R<sup>2</sup> = 0.764) and CNN (R<sup>2</sup> = 0.732) architectures. When pilot institutions adopted the model-guided curriculum adjustments across three vocational colleges, graduate employment rates rose by 15.3% and industry satisfaction scores increased by 21.6%, though these gains should be interpreted as observed associations rather than direct causal effects of the model itself. This work advances evidence-based educational planning and demonstrates practical viability for aligning vocational training with the shifting demands of the digital economy, while contributing to broader workforce development goals under current policy frameworks.</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

LSTM-CNN hybrid model for E-commerce talent demand prediction and intelligent program optimization in vocational colleges under the double first-class initiative

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月3日

查看全文
  • Ji Zhao

摘要

This study develops an intelligent e-commerce talent demand prediction system built on a hybrid Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) architecture, designed to support vocational college program optimization under China’s Double First-Class Initiative. The hybrid model pairs LSTM networks for temporal sequence modeling with CNN components for spatial feature extraction, processing multi-dimensional data drawn from job platforms, industry reports, and employment statistics. Experimental validation confirms that the hybrid approach outperforms traditional methods, with a 32.4% reduction in MAE, a 28.7% improvement in RMSE, and an R2 coefficient of 0.891—compared to standalone LSTM (R2 = 0.764) and CNN (R2 = 0.732) architectures. When pilot institutions adopted the model-guided curriculum adjustments across three vocational colleges, graduate employment rates rose by 15.3% and industry satisfaction scores increased by 21.6%, though these gains should be interpreted as observed associations rather than direct causal effects of the model itself. This work advances evidence-based educational planning and demonstrates practical viability for aligning vocational training with the shifting demands of the digital economy, while contributing to broader workforce development goals under current policy frameworks.