本論文主要研究以類神經網路模式預測季節性時間序列之有效性。利用適 當地建構樣本訓練集,網路經訓練後可作為季節性時間序列之預測工具。 文中亦提出移動學習法以期提高預測之準確度。並以台灣地區每季進口商 品與勞務總值則作為實證之研究。此季節性時間序列因受離群值之影響而 增加其預測困難度。實證結果顯示類神經網路模式之預測表現較傳統之統 計方法優異,即使此序列受到離群值之干擾。 / We investigate the effectiveness of neural networks for predicting the future behavior of seasonal time series. Utilizing the training set constructed properly, we can train the network who can be used to predict the future of seasonal time series. A shifting-learning method is also employed in order to obtained a better forecasting performance. The quarterly imports of goods and services of Taiwan between the first quarter of 1968 and the fourth quarter of 1990 are studied in the research. The series are contaminated with outliers, which will increase the difficulty of forecasting. Empirical results exhibit that neural networks model free approach have better prediction performance than the classical Box-Jenkins approach, even the series are contaminated with outliers.

本文將模糊理論的觀念，應用到時間數列分析上。研究重點包括模糊自相似度的定義與度量，模糊自迴歸係數的分析，模糊相似度辨識與自迴歸階次認定、模糊時間數列模式建構與預測等。我們首先給定模糊時間數列模式的概念與一些重要性質。接著提出模糊相似度的定義與度量，以及模式建構的流程。經由系統性的模擬與分析，我們建立階次認定的演算法則與認定程序。藉著詳細的演算比較這些類型的模糊時間數列。並以模糊關係方程式推導，提出合適的模糊時間數列模式建構方法。並利用提出的方法對台灣的景氣對策信號，及台灣結婚率建立模糊時間數列模式。最後，使用所建構的模糊時間數列模式對未來進行預測，以驗證所建構模糊時間數列模式的效率性與實用性。 / In modeling a time series the accuracy of various model constructions and forecasting techniques, certain rules and models are adhered to. Traditional methods on the model construction for a time series are based on the researchers' experience by choosing a "good" model, which will satisfactorily explain its dynamic behavior, from a model-base. But a fundamental question that often arises is: does the data exhibit the real case honestly? In this research we show how fuzzy time series construction be applied for this purpose. An order detection process for fuzzy time series is presented. Simulation has been used extensively to explore general properties of statistical procedures, and the approach is particularly useful in fuzzy time series construction. Statistical strategies typically consist of sequences of rules used repeatedly on the same data set. This paper is organized as follows: In Chapter 2 we will discuss about the definition of fuzzy time series as well as certain important properties. In Chapter 3, We use the similarity comparison process to decide the order of a fuzzy time series. Simulations and analysis with the results about various types of autocorrelation are experienced in Chapter 4. Finally, we apply our methods to three empirical examples, Taiwan business cycle index, marriage rate and numbers of students enrollment in Chapter 5. Chapter 6 is the conclusion and the discussion of future researches.

Schwarz (1997)提出均數回復過程（Mean-Reverting Process, MR）捕捉能源價格的動態過程，而Lucia and Schwarz (2002)將此模型結合確定季節性函數，並推導出期貨價格封閉解。然而，能源價格常會因為未預期事件的發生而產生大幅度的變動，為了描述價格跳躍的現象，Clewlow and Strickland (2000)延伸Schwarz的模型提出均數回復跳躍擴散模型（Mean-reverting jump diffusion process, MRJD），此模型除了保留均數回復模型對能源價格會回復至長期水準的描述外，再加上跳躍項來描述價格的異常變動。而Cartea and Figueroa (2005)則同時考慮季節性和跳躍因子，並推導出期貨價格封閉解。另外，雖然台灣目前並非京都議定書所規範的國家，但環境保護是未來的趨勢，故在衡量能源產業財務風險時，除了考慮相關原料和產品，應考慮碳權交易之影響。為了探討財務風險管理在能源產業之應用，本文以煉油廠為例，將其表示成特定期貨部位的投資組合，並透過計算投資組合風險值來衡量煉油廠的財務風險。文中使用結合季節性的均數回復過程、均數回復跳躍擴散過程進行模型配適。實證結果顯示，均數回復跳躍擴散模型在回溯測試下表現最佳；另外，考慮碳權交易後會使得煉油廠的財務風險上升。 / Schwarz (1997) proposes the mean-reverting process (MR) to model energy spot price dynamics, and Lucia and Schwarz (2002) extend this model by including mean reversion and a deterministic seasonality. This model can capture the mean-reversion of energy price, but fail to account for the huge and non-negligible price movement in the market. Clewlow and Strickland (2000) extend Schwarz’s model to mean-reverting jump diffusion process (MRJD). Cartea and Figueroa (2005) present a model which captures the most importance characteristics of energy spot prices such as mean reversion, jumps and seasonality, and provide a closed-form solution for the forward. Although Taiwan is not the member of Kyoto Protocol, but Environmental Protection is a trend in the future. In order to measure the financial risk induced by energy industries, we should consider the effect of emission trading. In this paper, we discuss the implication of financial risk management in energy industries by analyzing the exposure of refinery which represented certain energy futures portfolios. We use MR and MRJD process with seasonality to model energy spot price dynamics, and calibrate the parameters to historical data. And, we consider the interaction of all of positions and calculate the Value-at-Risk of portfolios. The results show that among various approaches the MRJD presents more efficient results in back-testing, and emission trading poses additional risk factors which will increase the financial risk for refineries.