Manufacturing Test Data Analysis-On Comparing Different Classification Algorithms

Due to the circuit complexity and the number of parameters involved, test relationship of a Test Program (TP) might not be fully discovered. Traditionally, TP setup are defined based on the domain expertise and gathered experience of an engineer. Such judgment is time consuming and could be inefficient especially when new products and technologies are rapidly developed for the competing market. If the complexity of a TP increases, the undetected interrelationship among tests in a TP will also increase. In this paper, inferences are performed to a huge and complex TP using different classification algorithms, with the primary goal to discover potential test relationships in a fast and efficient way. The mining output can be used as a reference and basis for test engineers to improve TP setup or to reprogram test machine to replace current exhaustive test policy.