Software Failure Time Prediction Method Based on Correlation Vector Regression Estimation

Abstract

The invention discloses a software failure time forecasting method based on relevant vector quantity regression estimation. Software failure time and time data with the number of m before the software failure time are learnt to accordingly capture interior dependency relations between failure time, and therefore the forecasting method based on reliability of relevant vector machine software is built. Because small sampling characteristics of forecasting of reliability of the software are fully considered, a kernel function technology can overcome the situations that the number of observational variables is larger than the number of observation samples and multi-collinearity exists between variables, the condition that model overfitting generated by modeling methods such as neural networks can be avoided. In the new forecasting method, model parameters are automatically adjusted to adapt to dynamic changing of the failure process, and therefore self-adaptation forecasting of reliability of the software is achieved and adaptive capacity of a forecasting model of software failure is effectively improved.

Description

【Technical field】 [0001] The invention relates to a software reliability test and a method for predicting software failure time data in the next or a long time in the future. 【Background technique】 [0002] Software reliability refers to the probability that software will not fail within a specified time under specified conditions. When solving the problem of reliability prediction, traditional solutions reflect the philosophy of large-sample statistics, and are prone to problems such as over-learning and poor applicability. [0003] Statistical learning theory is based on a relatively solid theoretical foundation, which provides a unified framework for solving finite sample learning problems. It can incorporate many existing methods, and is expected to help solve many problems that were difficult to solve, such as neural network structure selection problems, local minimum point problems, etc. Relevance vector machine (relevancevectormachine, RVM) is a sparse Bayesian lear...

AI Technical Summary

Problems solved by technology

When solving the problem of reliability prediction, the traditional solution reflects the philosophy of large sample statistics, which is prone to problems such as over-learning and poor applicability.

Images