Real-World Challenges in Building Accurate Software Fault Prediction Models

Real-World Challenges in Building Accurate Software Fault Prediction ModelsDR. ÇAĞATAY ÇATALTUBITAK (Research Council of TURKEY)

Predictive Modelling and Search Based Software Engineering, London, UK, 24-25 October 2011

2

Outline Introduction

Dependable Software Systems Motivation

Challenging Issues Fault prediction with no fault data Fault prediction with limited fault data Noise detection on measurement datasets Practical tools (Eclipse plug-in) Cross company vs. Within-company fault prediction

Our Models A Systematic Review Study Conclusion

Dependable Systems

Are we successful in building dependable software systems?

Safety (not being harmful for environment) Security (ability to protect the privacy) Reliability (ability to perform its function for a period of time) Availability (ability to serve whenever needed)

3

4

1. BRITISH ATM PAYS DOUBLE ! 19 March 2008

• ATM pays out double the amount withdrawn

• Dozens of customers lined up in front of ATM

• This continued until ATM ran out of money at 8 p.m.

Hull, England

A Generous British ATM...

5

A Sainsburry’s spokesman said“ We do not know how much the machine paid out at the

moment but the matter is under investigation”

A customer said“ I joined the queue and when I finally got to the front I drew

out 200 pound but it gave me 400 pound. The statement said I only drew out 200 pound. I don’t know whether I will have to pay it back”

The police said “ Those who benefited could face charges, but only if the

company administering the machine complained”.

2. ATM Pays Out Double the Cash, 16 January 2009

6

3. Tesco machine pays double, 18 August 2009

7

4. Dundee cash machine, 20 January 2011

8

But what happens if an ATM malfunctions and pays out less than you asked for?

We need dependable systems !

9

10

Motivation

Project Managers ask several questions: “How can I get the code into production faster? What code should we refactor? How should I best assign my limited resources

to different projects? How do I know if code is getting better or

worse as time goes on?”

“Baseline Code Analysis Using McCabe IQ”

•Software Metrics •Software Fault Prediction

11

Example: gcc project

/trunk/gcc/fold-const.c http://gcc.gnu.org/viewcvs/trunk/gcc/fold-const.c?re

vision=135517&view=markup

fold_binary’s CC value is 1159 ! Security problems or faults can occur

12

Vulnerability Report – Fold_Binary Method

http://vulnerabilities.aspcode.net/14389/fold+binary+in+fold+const+c+in+GNU+Compiler+Col.aspx

CHAPTER 2: Challenging Issues

14

Software Fault Prediction Modeling

Previous Version

TrainingLearnt

Hypothesis

Predict Faults

SoftwareMetrics

SoftwareMetrics

Current Project

Kn

ow

n F

ault D

ata

Un

kno

wn

Fau

lt Data

15

1. No Fault Data

SoftwareMetrics

SoftwareMetrics

Un

kno

wn

Fau

lt Data

LearntHypothesis

Previous Version

CurrentProject

Training

Predict Faults

Unknown Fault Data

* How does the software quality assurance team predict the software quality based on only the recorded software metrics?

- A new project type for organization- No quality measurement have not been collected

* Supervised learning approach cannot be taken

16

2. Limited Fault Data

SoftwareMetrics

SoftwareMetrics

Un

kno

wn

Fau

lt Data

LearntHypothesis

Previous Version

CurrentProject

Training

Predict Faults

Known Fault Data

Unknown Fault Data

* During decentralized software development, some companies may not collect fault data for their components* Execution cost of data collection tools may be expensive* Company may not collect fault data for a version due to the lack of budget- Can we learn both from labeled and unlabeled data?

17

3. Noise Detection Noisy modules degrades the performance of

machine learning based fault prediction models Attribute Noise Class Noise

Class noise impact classifiers more severely as compared to attribute noise

We need to identify noisy modules if they exist

Some cases: Developers may not report the faults Data entry and data collection errors

18

4. Practical Tools Earliest Work, Porter and Selby, 1990 .... Logistic Regression (Khoshgoftaar et al., 1999) Decision Trees (Gokhale et al., 1997) Neural Networks (Khoshgoftaar et al., 1995) Fuzzy Logic (Xu, 2001) Genetic Programming (Evett et al., 1998) Case-Based Reasoning (Khoshgoftaar et al.,

1997) Pareto Classification (Ebert, 1996) Discriminant Analysis (Ohlsson et al., 1998) Naive Bayes (Menzies et al., 2008) ...

Hundreds of research papers but lacking of practical tools…

19

5. Cross-Project vs. Within-Company Fault Prediction

Can we use cross-company (CC) data and predict the fault-proneness of program modules in the absence of fault labels?

CHAPTER 3: Models we built...

1. No Fault Data

21

22

1. No Fault Data Problem- Literature Zhong et al., 2004, Clustering and Expert based Approach

K-means and Neural Gas algorithms Mean vector and several statistical data such as min., max. Dependent on the capability of the expert

Zhong, S., T. M. Khoshgoftaar, and N. Seliya, “Unsupervised Learning for Expert-based Software Quality Estimation”, Proceedings of the 8th Intl. Symp. on High Assurance Systems Engineering, Tampa, FL, 2004, pp. 149-155.

23

1. No Fault Data Problem1. Our technique first applies X-means

clustering method to cluster modules and identifies the best cluster number.

2. The mean vector of each cluster is checked against the metrics thresholds vector. A cluster is predicted as fault-prone if at least one metric of the mean vector is higher than the threshold value of that metric.

[LOC, CC, UOp, UOpnd, TOp, TOpnd] [65, 10, 25, 40, 125, 70] (Integrated Software Metrics (ISM)

document)

24

• Datasets from Turkish white-goods manufacturer• Effective results are achieved• No expert opinion• Identification of threshold vector is difficult

2. Limited Fault Data Problem

25

26

2. Limited Fault Data Problem

We simulated small labeled-large unlabeled data problem with 5%, 10%, and 20% rates and evaluated the performance of each classifier under these circumstances.

Naive Bayes algorithm, even if it is a supervised learning approach, works best for small datasets

YATSI (Yet Another Two Stage Idea) improves the performance of Naive Bayes algorithm for large datasets if the dataset does not consist of noisy modules

We suggest Naive Bayes for limited fault data problem as well

3. Noise Detection

27

28

3. Noise DetectionOur hypothesis: A data object that has a non-faulty class label is considered a noisy instance if the majority of the software

metric values exceed their corresponding threshold values.

A data object that has a faulty class label is considered a noisy instance if all of the metric values are below their corresponding threshold values.

How to calculate software metrics threshold values?

R. Shatnawi, W. Li, J. Swain, T. Newman, Finding software metrics threshold values using ROC curves, Journal of Software

Maintenance and Evolution: Research and Practice 22 (1) (2010) 1–16.

29

How to Calculate Threshold Values The interval for the candidate threshold values is between

the minimum and maximum value of that metric in the dataset.

Shatnawi et al. (2010) stated that they chose the candidate threshold value that has the maximum value for both sensitivity and specificity, but such a candidate threshold may not always exist.

We calculated the AUC of the ROC curve that passes through three points, i.e., (0, 0), (1, 1), and (PD, PF), and we chose the threshold value that maximizes the AUC.

30

31

4. Practical Tools

32

33

4. Eclipse based Plug-in (RUBY)

Sample User Interfaces - Features

34

Result Views

35

5. Cross-Project Fault Prediction

36

37

5. Cross-Project Fault Prediction

We developed models based on software metrics threshold values If majority of software metrics thresholds values are

exceeded, the label of the module is faulty Otherwise, non-faulty label is assigned

Threshold values are calculated from the other projects (cross-company)

38

AUC, PD, PF

39

Results Case studies showed that the use of cross-company data is

useful for building fault predictors in the absence of fault labels and remarkable results are achieved.

Our threshold-based fault prediction technique achieved larger PD (but larger PF) value than Naive Bayes based approach.

For mission critical applications, PD values are more important than PF values because all of the faults should be removed before deployment.

In summary, we showed that cross-company dataset is useful.

40

4. Systematic Review

41

42

A Systematic Review Study

74 papers published between 1990 and 2007 27 journal papers 47 conference papers

We report distributions before and after 2005, since that was the year that the PROMISE repository was established.

Results The journals that published more than two fault model papers are:

IEEE Transaction of Software Engineering (9); Software Quality Journal (4); Journal of Systems and Software (3); Empirical Software Engineering (3)

14% of papers were published before 2000 and 86% after.

Types of data sets used by authors were: private (60%), partial (8%), public (31%), unknown (1%). ‘‘Partial” means data from open source projects that have not been circulated.

Since 2005 the proportion of private datasets has reduced to 31%, the proportion of public data sets has increased to 52%. There are 14% partial datasets and 3% unknown.

43

Results (cont’d)

Data analysis methods are machine learning (59%), statistics (22%), statistics and machine learning (18%) and statistics and expert opinion (1%).

After 2005 the distribution of methods is machine learning (66%), statistics (14%), statistics and machine learning (17%) and statistics and expert opinion (3%).

60% of papers used method level metrics, 24% used class level metrics, 10% were file level metrics, other categories less than 5%. 2005, 53% were method level, 24% were class level and 17% were file level (others less than 3%).

44

Suggestions

More studies should use class-level metrics to support early prediction.

Fault studies should use public datasets to ensure results can be repeatable and verifiable.

Researchers should increase usage of machine learning techniques.

45

46

Conclusion & Future Work Software fault prediction is still challenging and quite useful We need practical tools Prediction models can be used to predict vulnerability-prone

modules Challenges

How to make fault prediction work across projects ? How to build models when there is no fault data? How to build models when there is very limited fault data? How to remove noisy modules from datasets?

47

THANK YOU

Cagatay CATAL, Ph.D.cagatay.catal@bte.tubitak.gov.tr

www.cagataycatal.com