Abstract
This study focuses on the more efficient packaging reliability prediction by considering cluster analysis and regression algorithm simultaneously. The Wafer Level Chip Scale Packaging (WLCSP) experiencing Accelerated Thermal Cycling Test (ACTC) is observed. After confirming what the failure situation is, database with various dimensions is built through validated finite element models. Next, machine learning technique is introduced. One of algorithms, Polynomial Regression(PR), is selected to predict the reliabilities of different packaging because of its accuracy and advantage in calculation time. Moreover, that combining K-Means analysis obtains optimal result is the goal.