Stochastic Software Reliability: Modeling of Software Failures

Abstract

Given a test history consisting of record of times of software failures, together with a record of both the time at which each failure occurred and the type of that failure: 1) How many faults of each type remain in the software 2) How much added time on test is required to uncover a pre-specified number of faults? 3) If testing continues for a given increment of time, how many faults of each type will be uncovered? Stochastic models of software failures composed of a super population process that generates elements of a finite population of elements that are then successively sampled is used to construct both Bayesian and non Bayesian methods of parameter and predictive inference. Predictive validation of these models is done using NASA/GODDARD Software Engineering Lab data.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 15, 1996
Accession Number
ADA329608

Entities

People

  • Gordon M. Kaufman

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Acceptance Tests
  • Bayes Theorem
  • Bayesian Networks
  • Computations
  • Computing-Related Activities
  • Engineering
  • Estimators
  • Markov Chains
  • Maximum Likelihood Estimation
  • Models
  • Monte Carlo Method
  • Operations Research
  • Probabilistic Models
  • Reliability
  • Sampling
  • Software Development
  • Validation

Fields of Study

  • Computer science
  • Engineering

Readers

  • Computational Modeling and Simulation
  • Computer Science/Computer Engineering/Data Science/Digital Signal Processing.
  • Regression Analysis.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference