Test Database Analysis - Inferences from a Disposition Tree

Test Program (TP) is designed to have a few properties including test coverage, test flow and test condition specification. At present, these properties are adjusted based on expert experiences and some statistical tools. Heavy reliance on experts' opinions and gathered experiences on improving testing efficiency can be reduced if the structure of the test flow is well understood. This paper suggests a way to perform inferences in a very large and complex test database in order to assist test engineers to improve testing efficiency. The TP complexity can be reduced by incorporating (i) expert rules, (ii) test results, and (iii) heuristics to the database. We propose to perform inferences on constructed trees i.e. source tree and Disposition Tree (DispTree) by counting path frequency and to provide insights for test engineers to identify redundant and misplaced test in a test flow. The proposed approach is evaluated on a considerably huge and complex TP with up to a thousand tests for dual-core microchip.