轉折型時間序列在現實生活中常常可見，例如因戰爭、政策改變、罷工或自然界的條件劇變等，而使時間序列的走勢發生明顯的轉變。傳統上，對這種轉折型時間序列資料進行轉折點的偵測時，大部分均從事後的觀點，主觀上先行認定結構轉變發生的時點，而後再以檢定加以確認。但此種方法過於主觀，而且轉型並非一蹴可幾，若以單一的轉折點來解釋轉型的現象，似乎不太恰當。 　　有鑑於此，本文利用模糊轉折區間統計認定法，以事前的觀點，對具有平均數或變異數改變的轉折型時間序列進行轉折區間的認定。並以匯率及貿易餘額的實際例子，利用我們所提出的方法進行單變數及雙變數的模糊分類，進而求出個別及聯合的轉折區間。 / Structure-changing time series are often seen in daily life. For example, war, change of policy, labor strike, or change of natural phenomena result in obvious change of time series. Most of detection of change points for structure-changing time series take place afterwards. In this paper, we pre-sent a change periods detection method for trend time series using the concept of fuzzy logic. Empirical example about exchange rate and balance of international trade is illustrated with detailed analysis.

近年來，非線性時間數列分析是一個快速發展的課題，其中最為人所矚目 的是門檻模式。從過去許多文獻得知，一個簡單門檻模式對於某些型態時 間數列的描述，如結構性改變的行為趨勢，比一般線性ARMA模式更能解釋 實際情況。在本篇論文中，我們將討論有關門檻模式及結構性改變分析的 問題。對於模式的建立，我們提出一個轉型期的觀念，替代傳統尋求一個 轉捩點的方法，進而提出一個結構性改變ARIMA模式有效建立的程序。最 後，我們以台灣出生率當作應用分析的範例，並且利用建立的結構性改變 ARIMA模式，及其他傳統門檻TAR模式，傳統線性分析方法等進行預測分析 及比較。 / Non-linear time series analysis is a rapidly developing subject in recent years. One of special families of non-linear models is threshold model. Many literatures have shown that even simple threshold model can describe certain types of time series, such as structural change behavior, more faithful than using linear ARMA models. In this paper, we discuss some problems about the threshold model and structural change analysis. Instead of finding the change point, we present the change period concepts on the model- building. An efficient algorithem on constructing the structure change ARIMA models is proposed. Finally, we demonstrate an example about the birth rate of Taiwan, and the comparison of forecasting performance for the structure change ARIMA model with alternative models are also made.

由於許多經濟指標的定義不明確，或是因為資料蒐集的時間不一，導致代表經濟景氣的數值，實際上即具有相當大的的不確定性。傳統的方法多不考慮這樣的模糊性,而傾向尋找一準確的模式轉折點。本文則以模糊數學的方法,運用模糊分類法以及模糊熵，訂定一個評判的準則。藉以找出一時間數列模式發生變化的轉折區間。最後以台灣經濟景氣指標為例，說明此方法可不需對資料的模式有任何事先的認知，即可得出與傳統方法相近，甚至更為合理的預測結果。 / Unlike conventional change points detecting, which seeks to find a decision boundary between classes for certain structural changed time series, the purpose of this research is to investigate a new approach about fuzzy change period identification. Based on the concept of fuzzy theory, we propose a procedure for the - level of fuzzy change period detecting and prove some useful properties for a fuzzy time series. We use some numerical examples to demonstrate how these procedures can be applied. Finally, experimental results show that the proposed detecting approach for structure change of fuzzy time series is available and practical in identifying the alpha-level of fuzzy change period.

Les logiciels sont en constante évolution, nécessitant une maintenance et un développement continus. Ils subissent des changements tout au long de leur vie, que ce soit pendant l'ajout de nouvelles fonctionnalités ou la correction de bogues. Lorsque les logiciels évoluent, leurs architectures ont tendance à se dégrader et deviennent moins adaptables aux nouvelles spécifications des utilisateurs. En effet, les architectures de ces logiciels deviennent plus complexes et plus difficiles à maintenir à cause des nombreuses dépendances entre les artefacts. Par conséquent, les développeurs doivent comprendre les dépendances entre les artefacts des logiciels pour prendre des mesures proactives qui facilitent les futurs changements et ralentissent la dégradation des architectures des logiciels. D'une part, le maintien d'un logiciel sans la compréhension des les dépendances entre ses artefacts peut conduire à l'introduction de défauts. D'autre part, lorsque les développeurs manquent de connaissances sur l'impact de leurs activités de maintenance, ils peuvent introduire des défauts de conception, qui ont un impact négatif sur l'évolution du logiciel. Ainsi, les développeurs ont besoin de mécanismes pour comprendre comment le changement d'un artefact impacte le reste du logiciel. Dans cette thèse, nous proposons trois contributions principales : La spécification de deux nouveaux patrons de changement et leurs utilisations pour fournir aux développeurs des informations utiles concernant les dépendances de co-changement. La spécification de la relation entre les patrons d'évolutions des artefacts et les fautes. La découverte de la relation entre les dépendances des anti-patrons et la prédisposition des différentes composantes d'un logiciel aux fautes. / Program maintenance accounts for the largest part of the costs of any program. During maintenance activities, developers implement changes (sometimes simultaneously) on artefacts to fix bugs and to implement new requirements. Thus, developers need knowledge to identify hidden dependencies among programs artefacts and detect correlated artefacts. As programs evolved, their designs become more complex over time and harder to change. In the absence of the necessary knowledge on artefacts dependencies, developers could introduce design defects and faults that causes development and maintenance costs to rise. Therefore, developers must understand the dependencies among program artefacts and take proactive steps to facilitate future changes and minimize fault proneness. On the one hand, maintaining a program without understanding the different dependencies between their artefacts may lead to the introduction of faults. On the other hand, when developers lack knowledge about the impact of their maintenance activities, they may introduce design defects, which have a negative impact on program evolution. Thus, developers need mechanisms to understand how a change to an artefact will impact the rest of the programs artefacts and tools to detect design defects impact. In this thesis, we propose three principal contributions. The first contribution is two novel change patterns to model new co-change and change propagation scenarios. We introduce the Asynchrony change pattern, corresponding to macro co-changes, i.e., of files that co-change within a large time interval (change periods), and the Dephase change pattern, corresponding to dephase macro co-changes, i.e., macro co-changes that always happen with the same shifts in time. We present our approach, named Macocha, and we show that such new change patterns provide interesting information to developers. The second contribution is proposing a novel approach to analyse the evolution of different classes in object-oriented programs and to link different evolution behaviour to faults. In particular, we define an evolution model for each class to study the evolution and the co-evolution dependencies among classes and to relate such dependencies with fault-proneness. The third contribution concerns design defect dependencies impact. We propose a study to mine the link between design defect dependencies, such as co-change dependencies and static relationships, and fault proneness. We found that the negative impact of design defects propagate through their dependencies. The three contributions are evaluated on open-source programs.

Les logiciels sont en constante évolution, nécessitant une maintenance et un développement continus. Ils subissent des changements tout au long de leur vie, que ce soit pendant l'ajout de nouvelles fonctionnalités ou la correction de bogues. Lorsque les logiciels évoluent, leurs architectures ont tendance à se dégrader et deviennent moins adaptables aux nouvelles spécifications des utilisateurs. En effet, les architectures de ces logiciels deviennent plus complexes et plus difficiles à maintenir à cause des nombreuses dépendances entre les artefacts. Par conséquent, les développeurs doivent comprendre les dépendances entre les artefacts des logiciels pour prendre des mesures proactives qui facilitent les futurs changements et ralentissent la dégradation des architectures des logiciels. D'une part, le maintien d'un logiciel sans la compréhension des les dépendances entre ses artefacts peut conduire à l'introduction de défauts. D'autre part, lorsque les développeurs manquent de connaissances sur l'impact de leurs activités de maintenance, ils peuvent introduire des défauts de conception, qui ont un impact négatif sur l'évolution du logiciel. Ainsi, les développeurs ont besoin de mécanismes pour comprendre comment le changement d'un artefact impacte le reste du logiciel. Dans cette thèse, nous proposons trois contributions principales : La spécification de deux nouveaux patrons de changement et leurs utilisations pour fournir aux développeurs des informations utiles concernant les dépendances de co-changement. La spécification de la relation entre les patrons d'évolutions des artefacts et les fautes. La découverte de la relation entre les dépendances des anti-patrons et la prédisposition des différentes composantes d'un logiciel aux fautes. / Program maintenance accounts for the largest part of the costs of any program. During maintenance activities, developers implement changes (sometimes simultaneously) on artefacts to fix bugs and to implement new requirements. Thus, developers need knowledge to identify hidden dependencies among programs artefacts and detect correlated artefacts. As programs evolved, their designs become more complex over time and harder to change. In the absence of the necessary knowledge on artefacts dependencies, developers could introduce design defects and faults that causes development and maintenance costs to rise. Therefore, developers must understand the dependencies among program artefacts and take proactive steps to facilitate future changes and minimize fault proneness. On the one hand, maintaining a program without understanding the different dependencies between their artefacts may lead to the introduction of faults. On the other hand, when developers lack knowledge about the impact of their maintenance activities, they may introduce design defects, which have a negative impact on program evolution. Thus, developers need mechanisms to understand how a change to an artefact will impact the rest of the programs artefacts and tools to detect design defects impact. In this thesis, we propose three principal contributions. The first contribution is two novel change patterns to model new co-change and change propagation scenarios. We introduce the Asynchrony change pattern, corresponding to macro co-changes, i.e., of files that co-change within a large time interval (change periods), and the Dephase change pattern, corresponding to dephase macro co-changes, i.e., macro co-changes that always happen with the same shifts in time. We present our approach, named Macocha, and we show that such new change patterns provide interesting information to developers. The second contribution is proposing a novel approach to analyse the evolution of different classes in object-oriented programs and to link different evolution behaviour to faults. In particular, we define an evolution model for each class to study the evolution and the co-evolution dependencies among classes and to relate such dependencies with fault-proneness. The third contribution concerns design defect dependencies impact. We propose a study to mine the link between design defect dependencies, such as co-change dependencies and static relationships, and fault proneness. We found that the negative impact of design defects propagate through their dependencies. The three contributions are evaluated on open-source programs.