臺灣水果價格之預測 ─ 機器學習之應用

Abstract

臺灣自古享有“水果王國”的美譽。然而，臺灣的水果價格波動較大。了解未來的價格趨勢和影響價格的因素，將為買家和賣家以及政策制定者提供更大的保證，並有助於製定有效的政策來減輕變化。本研究以臺灣前三大出口水果釋迦、鳳梨、芒果為研究對象，建立相應的價格預測模型。該研究涵蓋臺灣北部、中部和南部三個市場。本研究使用LSTM（長短期記憶）模型，對2002年至2021年的價格數據進行預測。結果表明，該模型準確捕捉了價格趨勢的波動，鳳梨和芒果的價格預測更加準確。本研究展示了預測農作物價格的一個有前景的工具。

Taiwan has enjoyed the reputation of “fruit kingdom” for a historically long time. However, the prices of fruits in Taiwan fluctuate greatly. Understanding the future price trends and factors influencing prices, it would provide greater assurance for both buyers and sellers, as well as policymakers, and help inform effective policies mitigating variation. This study focuses on the top three exporting fruit in Taiwan, pineapple, custard apple, and mango, and establishes corresponding price prediction models. The research covers three markets in northern, central, and southern Taiwan. This research uses the LSTM (Long Short-Term Memory) model, to predict price data for 2002 to 2021.Results show that the models accurately capture the fluctuations of price trend. Price prediction for pineapple and mango are more accurate. This study show case a promising tool for predicting price of agricultural crops.