運用具單變換點雙參數廣義柏拉圖模型於開放源碼軟體之錯誤分佈分析

Applying 2-Parameter Generalized Pareto Model with Single Change-Point to Analyze the Fault Distribution of Open Source Software

欒紹蒲(Shao-Pu Luan)

指導教授 ： 黃慶育

國立清華大學/電機資訊學院/資訊工程學系/碩士(2012年)

https://doi.org/10.6843/NTHU.2012.00514

摘要

普遍性同意開放源碼軟體在現代社會上日益扮演一個重要的角色。儘管文獻已經廣泛探討過開放源碼軟體的可靠度議題，但是對開放源碼軟體的錯誤分佈研究則相對缺乏。在過去有一些研究發現指出Pareto principle，traditionally-used Pareto distribution以及Weibull distribution模型均可以用來描述軟體的錯誤分佈；然而在我們前面的研究中發現，一個與Pareto principle相關的2-parameter generalized Pareto distribution (2-GPD) 模型能更有益於塑造軟體的錯誤分佈。
本文探討一種基於2-GPD為修改的模型：2-parameter generalized single change-point Pareto distribution (SCP-2GPD)，其變換點的選擇與Pareto principle有相當大的關係。此研究聚焦在塑造開放源碼軟體的錯誤分佈，並且呈現一些SCP-2GPD模型的數學性質。研究資料來源基於Apache以及Mozilla均被施行 (勘查於名為Bugzilla的開放源碼軟體錯誤資料庫)，以確認SCP-2GPD模型的錯誤分佈之預測能力。與其他相關錯誤分佈模型比較後，本研究結果指出提議的SCP-2GPD模型在開放源碼軟體的錯誤分佈具有相當準確之預測能力。這些發現對於分析真實各種開放源碼軟體的錯誤分佈都有很大的啟發。

關鍵字

開放源碼軟體 ； 錯誤分佈 ； 柏拉圖法則 ； 韋伯分佈模型 ； 變換點

並列摘要

There is a general agreement that open source software (OSS) plays an increasingly critical role in the modern society. In the past, some research findings showed that the Pareto principle, the traditionally-used Pareto distribution (PD) and the Weibull distribution (WD) models would be able to describe the distribution of software fault; a Pareto principle related 2-parameter generalized Pareto distribution (2-GPD), however, could be more useful to model the distribution of software faults in our previous study. This paper studies a modification of the 2-GPD model called the 2-parameter generalized single change-point Pareto distribution (SCP-2GPD) model, and the selection of change-point is highly pertinent to the Pareto principle. The research focuses on modeling the distribution of OSS faults, and some mathematical properties of the SCP-2GPD model are presented. Sources of data based on Apache and Mozilla found in bug database of OSS called Bugzilla are performed in order to ascertain the prediction capability of fault distribution for the SCP-2GPD model. Compared with other fault distribution models, the findings suggest that the proposed SCP-2GPD model has a fairly accurate prediction capability of fault distribution of OSS. These findings have implications for analyzing the fault distribution of real-life-various OSS.

並列關鍵字

Open source software ； fault distribution ； Pareto principle ； Weibull distribution model ； change-point

參考文獻

[1] T. J. Ostrand, E. J. Weyuker, and R. M. Bell, “Predicting the Location and Number of Faults in Large Software Systems,” IEEE Trans. on Software Engineering, Vol. 31, No. 4, pp. 340-355, April 2005.

[3] H. Zhang “On the Distribution of Software Faults,” IEEE Trans. on Software Engineering, Vol. 34, No. 2, pp. 301-302, March/April 2008.

[4] M. R. Lyu, Handbook of Software Reliability Engineering, McGraw Hill, 1996.

[5] C. Y. Huang, M. R. Lyu, and S. Y. Kuo, “A Unified Scheme of Some Non-Homogenous Poisson Process Models for Software Reliability Estimation,” IEEE Trans. on Software Engineering, Vol. 29, No. 3, pp. 261-269, March 2003.

[6] Y. Zhou and J. Davis, “Open source software reliability model: an empirical approach”, Proceedings of the 5th Workshop on Open Source Software Engineering, pp. 1-6, May 2005.