- 學位論文
用於軟體除錯之修正相似係數頻譜式錯誤定位方法之分析
Analysis of Spectrum-Based Fault Localization Methods with Modified Similarity Coefficients for Software Debugging
摘要
在軟體錯誤定位技術的相關研究中，近年來以頻譜式(spectrum-based)的方法被最為廣泛地探討。這是因為其具有能自動分析、易於實行以及有效地定位出錯誤位置的特性。而大部分頻譜式的軟體錯誤定位技術主要仰賴於相似係數(similarity coefficients)的使用。然而，現存的相似係數並不能適當地反映出在軟體錯誤定位中，成功與失敗的測試案例間對於錯誤發生的位置所能提供的資訊量是不相等的。對使用於錯誤定位的相似係數而言，失敗的測試案例應該要有比成功的測試案例更大的影響力。 因此為了更精確地評估相似係數在計算成功與失敗的測試案例間的平衡，本研究針對幾種廣泛應用於軟體錯誤定位的相似係數，提出一套加入了權重機制的改良模型。並透過使用合併數種涵蓋標準(combination of multiple coverage types)的技術來實行以及評估我們所提出的改良相似係數模型在軟體錯誤定位上的表現。研究的結果指出：我們所提出的相似係數改良模型在軟體錯誤定位上，皆有較使用原方法更佳的效率，即使是在近來相關研究中表現最好的Ochiai係數也不例外，因此加入了權重的相似係數方法的確能更有效地定位軟體中的錯誤。
並列摘要
In the current researches of fault localization, the spectrum-based techniques are frequently and widely studied for researchers since it can automatically and effectively localize the faults of software and also can be implemented easily. Most of the spectrum-based fault localization techniques rely on using the similarity coefficients. However, the existing similarity coefficients for fault localization lack a measure to properly reflect the relation between the failing and passing test cases. The failing test cases are expected to provide more information to the similarity coefficients than the passing test cases. In order to accurately evaluate the relative importance of failing and passing test cases in the similarity coefficients, several widely used similarity coefficients in fault localization are investigated in our study. The modified similarity coefficients which are assigned the weight of the failing/passing test cases are proposed and analyzed by using the multiple coverage-combined technique. The performance evaluation of our approach indicates that the proposed approach performs better than the original methods including the Ochiai coefficient, which has the best performance in current researches. Therefore, our study shows that assigning the weight of failing and passing test cases to the similarity coefficients can localize the faults more effectively.