Development of software reliability models using a hybrid approach and validation of the proposed models using big data

This article proposes three software reliability models based on a hybrid approach combining NHPP models, Weibull model, and exponential model. The software failure is first categorised into three categories, namely pure software failures, hardware-induced software failures, and user-induced softwar...

Full description

Saved in:
Bibliographic Details
Published inThe Journal of supercomputing Vol. 76; no. 4; pp. 2252 - 2265
Main Authors Govindasamy, P., Dillibabu, R.
Format Journal Article
LanguageEnglish
Published New York Springer US 01.04.2020
Springer Nature B.V
Subjects
Big Data
Compilers
Computer Science
Data analysis
Failure
Failure analysis
Genetic algorithms
Hardware
Interpreters
Maximum likelihood estimates
Maximum likelihood estimation
Model accuracy
Parameter estimation
Processor Architectures
Programming Languages
Reliability analysis
Software reliability
Hardware
Parameters
Big data analysis
Reliability models
Software failures
Estimation accuracy
Online AccessGet full text

Cover

More Information
Summary:This article proposes three software reliability models based on a hybrid approach combining NHPP models, Weibull model, and exponential model. The software failure is first categorised into three categories, namely pure software failures, hardware-induced software failures, and user-induced software failures. Based on the failure behaviour, NHPP models were adapted for pure software failures, Weibull model for hardware-induced failures, and exponential model for user-induced software failures. The failure intensity function, mean value function, and reliability function were determined. The proposed models are validated using big data analysis. From the data collected during the testing phase, the optimal values of parameters were estimated using maximum likelihood estimation and genetic algorithm. The expected number of failures and the cumulative number of failures were calculated, and comparison was made between the observed values to show the performance of the proposed models. A comparison criterion was also proposed to confirm the estimation accuracy. Finally, a t test was conducted to test the significance of the difference between the observed and estimated values. Experimental results confirm the better estimation accuracy of the proposed models.
ISSN:0920-8542
1573-0484
DOI:10.1007/s11227-018-2457-8