應用氣象資料配合類神經網路建立茼蒿需求量預測模式──以台北第一果菜批發市場為例

Establishing a Demand Forecast Model for Crowndaisy Chrysanthemum with Weather Data and Neural Network--An Empirical Example of Taipei’s First Vegetable Wholesale Market

由於直覺上火鍋與每日天氣的變化以及假日狀況有關，而茼蒿乃是吃火鍋時大家最常想到的蔬菜，故本文主要目的在了解利用氣象資料建立茼蒿每日需求量預測模式之可行性。類神經網路乃是利用人工神經元所組成的網路模式來模仿生物神經網路的高度學習能力，在其他領域應用廣泛，且其中的倒傳導式類神經網路模式在預測應用上有很好的成效。因此本文利用類神經網路來建立預測模式。本文以實驗法得出最適的類神經網路預測架構之隱藏層為兩層，每層各一個處理單元，學習速率為0.6，學習循環為5000次，收斂之RMS為0.0888873。經由各項類神經網路常用的驗證方式驗證後，發現以氣象資料配合應用類神經網路來建立茼蒿每日需求量預測模型，在實際操作上是可行的。 In Taiwan, Crowndaisy Chrysanthemum(CC) is a famous ingredient of chafer and eating chafer is generally known as being related to weather and holiday conditions. The purpose of this paper is to study the feasibility of establishing a daily demand forecast mode of CC with weather data. As for methodology adopting, backpropagation, a neural network mode, is introduced to solve this problem because of its admirable learning ability and its successful application in many research fields. An experimentation is used to determine and select out the framework of the neural network for forecasting. The selected framework for forecasting in this paper contains two hidden layers, one neuron in each hidden layer, 0.6 as its learning rate, and 5,000 to be its learning cycles. The convergence RMS value is 0.0888873. After verifying the selected network with several general validate methods of neural network, we conclude that it is feasible to apply neural network to establish a daily demand forecast model of CC with weather data.