Optimizing Regression Testing Through Defect History Code Risk and Test Coverage Analytics

Abstract

Continuous integration turns regression testing into a per-change decision: which tests to run, in what order, within a fixed time budget, and at what risk of letting a regression escape. This review examines how four evidence families, defect and failure history, code change risk, dependency reach, and coverage evidence, should be combined for that decision. A structured integrative protocol searched six bibliographic sources for work published up to 2025 on regression test selection and test case prioritization, and appraised studies for fault realism, time-ordered evaluation, cost reporting, and treatment of flaky tests. The synthesis shows that each family fails in a characteristic way. Failure history favors already broken and flaky tests and is blind to new code; churn metrics locate defect-prone files but not the test that reveals the fault; static dependency analysis supports safe selection at the price of low precision; and coverage maps decay as code evolves. Cost-blind metrics such as APFD can rate a schedule highly while it consumes most of the budget, so cost-cognizant APFDc and budget-truncated detection are the preferred criteria. The article proposes the D3C framework, which estimates a calibrated detection probability for each test, orders tests by probability per unit of execution cost, and protects safety through a mandatory floor and an exploration quota. Six falsifiable propositions and a staged validation design are specified.