Software Bug Localization developed a revolutionary Information Retrieval (IR) model. It points software development teams to the source of their coding errors so they can fix them up to 80% faster than with existing solutions.
Complex software projects create many places for errors or defects. The more layers you add, the more chances for bugs. Sifting through pages of data to find that bad code can take hours— even weeks. How much time does your company spend chasing down bugs? What does it cost in money and manpower?
Imagine what you could do with that freed-up time! Develop new cutting edge software. Advance from concept to product-ready software in a fraction of the time. Get a speed advantage over the competition.
And what about the financial advantage? You will decrease the overall cost and risk of software development, testing and migration. Think of the boost to bottom line profits.
The technology uses version histories of a software project to estimate the probability for a defect to be associated with a given version of the project. The IR framework holds each of these priors as well as a defect history. These are checked to determine the posterior probability of the file being the cause of the bug.
Just including the version and defect histories in the IR framework improved the mean average precision (MAP) for bug localization by 30%.
The model also includes a temporal decay in the estimation of the priors. When the time decay is factored in with the estimates of the priors and the defect history, the MAP improved up to 80%.