- 學位論文
毛豬價格預測：馬可夫狀態轉換模型之應用
Hog Price Forecasting:An Application of Markov-Switching Model
摘要
毛豬一直都是台灣極為重要的產業，在國人的日常飲食消費中，豬肉亦佔了極大的比例，所以毛豬價格的漲跌對於國人的影響甚劇，因此有效掌握毛豬價格的波動極為重要。根據過去的研究結果發現，非線性時間數列模型對於具有結構性改變的資料，其模型配適與預測能力皆比線性模型佳，其中馬可夫狀態轉換模型，相較於其他的非線性模型又更能捕捉到此類資料的特性。有鑑於此，本研究採用馬可夫狀態轉換自我迴歸模型(MS-(V)AR Model)來進行毛豬產地價格、豬腹協肉零售價格、豬腹協肉進口價格、進口玉米(飼料用)農場價格資料的分析及預測，並使用均方根差(RMSE)做為預測能力評估的準則。 研究結果發現，在單變量模型中，毛豬產地價格以MSM(3)-AR(6)為最適模型，豬腹協肉零售價格以MSMH(3)-AR(3)為最適模型，豬腹協肉進口價格以MSM(3)-AR(1)為最適模型，進口玉米(飼料用)農場價格以MSMH(3)-AR(1)為最適模型；而在多變量模型中，四種價格資料以MSMH(3)-VAR(2)模型來進行配適最適合。在預測能力方面，整體而言，以多變量馬可夫狀態轉換自我迴歸模型作為未來價格波動的最佳預測模型。
並列摘要
Hog industry plays an important role in Taiwan agricultural sector, while pork is also major meat consumption in Taiwan. Therefore, it’s very important to forecast related prices of hog. Pervious studies have show that Markov switching model has a better performance than other nonlinear time series models on forecasting for the time series data with structural changes. The study thus uses the Markov switching model to analyze and forecast hog related prices, farm gate prices of hog, retail prices and import prices of pork belly, and import prices of feeding used corn. The comparisons between univariate Markov-switching model and multivariate Markov-switching model is also conducted based upon the RMSE values. The study results of univariate models show that MSMH(3)-AR(6) model is the most suitable model on forecasting hog farm gate prices, MSMH(3)-AR(3) is the most suitable model on forecasting retail prices of pork belly, MSMH(3)-AR(1) the most suitable model on forecasting import prices of pork belly, and MSMH(3)-AR(1) is the most suitable for forecasting import prices of feeding used corn. Moreover, the study results of multivariate model also show that MSMH(3)-VAR(2) has a better performance than that of others on forecasting the four hog related prices. According to the study results, the multivariate Markov-switching model is thus suggested as a forecasting model for future predict on hog related prices.