基于遷移學(xué)習(xí)的軟件缺陷預(yù)測研究與實現(xiàn)

13800字
自己原創(chuàng)的畢業(yè)論文，已經(jīng)通過校內(nèi)系統(tǒng)檢測，重復(fù)率低，僅在本站獨家出售，大家放心下載使用

摘要 軟件缺陷預(yù)測研究通常使用公司內(nèi)部數(shù)據(jù)建模，極少數(shù)使用跨公司數(shù)據(jù)來訓(xùn)練預(yù)測模型。由于缺少目標(biāo)公司訓(xùn)練數(shù)據(jù)，使用這些建立于公司內(nèi)部數(shù)據(jù)的模型是困難的。近來，遷移學(xué)習(xí)吸引了越來越多的注意力，主要關(guān)于使用來自相關(guān)源領(lǐng)域的數(shù)據(jù)在目標(biāo)領(lǐng)域構(gòu)建分類器。當(dāng)訓(xùn)練分布和測試樣例不同時，遷移學(xué)習(xí)就很有用處，但是，它是否適用于跨公司軟件缺陷預(yù)測呢？
我們考慮源數(shù)據(jù)和目標(biāo)數(shù)據(jù)取自于不同公司的跨公司缺陷預(yù)測方案。為了利用跨公司數(shù)據(jù)，學(xué)者們嘗試通過開發(fā)遷移學(xué)習(xí)方法來構(gòu)建更快更高效的預(yù)測模型。本課題使用樸素貝葉斯（Naive Bayes）算法來完成對訓(xùn)練數(shù)據(jù)的分類，實現(xiàn)有效知識的遷移。該算法使用基于合適的特征屬性的所有訓(xùn)練樣本信息，通過給訓(xùn)練數(shù)據(jù)加權(quán)分類來評估測試數(shù)據(jù)的分布和跨公司數(shù)據(jù)的遷移。本文呈現(xiàn)了相關(guān)方法的理論分析，并實現(xiàn)了對訓(xùn)練數(shù)據(jù)的分類?？梢缘贸鼋Y(jié)論：當(dāng)只有極少數(shù)的本地訓(xùn)練數(shù)據(jù)用來訓(xùn)練分類器時，這些基于特征層面的不同分布的訓(xùn)練數(shù)據(jù)的實用信息是有幫助的。該遷移學(xué)習(xí)方法可帶來理想的資源分配策略，即減少軟件測試花費(fèi)以及增加軟件測試過程的高效性。

關(guān)鍵詞:遷移學(xué)習(xí) 軟件缺陷預(yù)測 分類器 跨公司數(shù)據(jù)

Research and implementation of software defect prediction based on transfer learning

Abstract ：Software defect prediction studies usually built models using within-company data, but very few focused on the prediction models trained with cross-company data. It is difficult to employ these models which are built on the within-company data in practice, because of the lack of these local data repositories. Recently, transfer learning has attracted more and more attention for building classifier in target domain using the data from related source domain. It is very useful in cases when distributions of training and test instances differ, but is it appropriate for cross-company software defect prediction?
In this paper, researchers consider the cross-company defect prediction scenario where source and target data are drawn from different companies. This topic using the naive Bias (Naive Bayes) algorithm to complete the classification of the training data, realize the transferring of effective knowledge . by using the information of all the proper features in training data. Our solution estimates the distribution of the test data, and transfers cross-company data information into the weights of the training data. On these weighted data, the defect prediction model is built.This article presents a theoretical analysis for the comparative methods, and shows the experiment results on the data sets. It is concluded that when there are too few local training data to train good classifiers, the useful knowledge from different-distribution training data on feature level may help. We are optimistic that our transfer learning method can guide optimal resource allocation strategies, which may reduce software testing cost and increase effectiveness of software testing process.

Keywords: Transfer Learning Software Defect Prediction Classifier Scientific Projects


目錄
第一章引言 1
1.1 研究動機(jī) 1
1.2 研究現(xiàn)狀 1
1.3 本文的主要工作 3
1.4 本文的組織結(jié)構(gòu) 4
第二章 需求分析與相關(guān)技術(shù)理論 5
2.1 需求分析 5
2.2 相關(guān)技術(shù) 5
2.2.1開發(fā)環(huán)境 5
2.2.2使用的分類算法 6
2.3遷移學(xué)習(xí)簡述 8
第三章 軟件缺陷預(yù)測與簡析 10
3.1軟件缺陷預(yù)測 10
3.2簡析 10
第四章 基于遷移學(xué)習(xí)的軟件缺陷預(yù)測 13
4.1 構(gòu)造樸素貝葉斯分類器 13
4.2相關(guān)訓(xùn)練集屬性 14
4.3代碼實現(xiàn) 17
第五章 實驗結(jié)果及算法復(fù)雜度分析 22
5.1 用于比較的算法 22
5.2算法復(fù)雜度分析及比較 22
5.3運(yùn)行結(jié)果 23
5.4小結(jié) 24
第六章 結(jié)論和展望 25
6.1結(jié)論 25
6.2展望 25
致謝 27
參考文獻(xiàn) 28