Scholarlink

The Taiwanese livestock industry grapples with escalating costs and challenges such as labor shortages, environmental changes, pollution, and disease threats. Trade liberalization and cheaper imports further highlight the need for industry modernization. In slaughter and food processing, predicting prices requires careful consideration of operational costs and pricing strategies. To analyze historical chicken meat prices, we compared ARIMA and LSTM models, common in time series forecasting. Additionally, we introduced STL-LSTM, combining STL and LSTM. These aimed for accurate real-time price predictions, with MSE used to measure accuracy. The study found that the LSTM model exhibited high predictive accuracy with the available data. Across various prediction horizons (1, 3, and 6 steps), the root mean squared error (RMSE) values were 3.09, 3.27, and 3.34, respectively, and the mean absolute percentage error (MAPE) values were 2.29, 2.95, and 2.92, respectively. These results surpassed those of the STL-LSTM and ARIMA models significantly. The LSTM model demonstrated capability in anticipating and signaling abnormal price fluctuations, offering valuable insights for pricing decisions within the industry.

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

Implementation on the Price Prediction and Analysis of Chicken Breast with Deep Learning Techniques

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Chao-Hung Liu,
  • Yu-Wei Chan,
  • Po-Yen Lin,
  • Chih-Hung Chang,
  • Chao-Tung Yang

摘要

The Taiwanese livestock industry grapples with escalating costs and challenges such as labor shortages, environmental changes, pollution, and disease threats. Trade liberalization and cheaper imports further highlight the need for industry modernization. In slaughter and food processing, predicting prices requires careful consideration of operational costs and pricing strategies. To analyze historical chicken meat prices, we compared ARIMA and LSTM models, common in time series forecasting. Additionally, we introduced STL-LSTM, combining STL and LSTM. These aimed for accurate real-time price predictions, with MSE used to measure accuracy. The study found that the LSTM model exhibited high predictive accuracy with the available data. Across various prediction horizons (1, 3, and 6 steps), the root mean squared error (RMSE) values were 3.09, 3.27, and 3.34, respectively, and the mean absolute percentage error (MAPE) values were 2.29, 2.95, and 2.92, respectively. These results surpassed those of the STL-LSTM and ARIMA models significantly. The LSTM model demonstrated capability in anticipating and signaling abnormal price fluctuations, offering valuable insights for pricing decisions within the industry.