The application of ReliefF algorithm for identifying CTQ in complex products

In the feature set of complex products which are in high dimension, the set usually contains useful information, irrelevant information and redundancy information. However, the former one is usually buried in the latter two. Therefore, the recognition of the most useful information in the original data set, which is defined as the identification of Critical-To-Quality features, becomes a key process in the field of quality control. The traditional methods include QFD, Taguchi loss function and Decision tree, etc. However, almost none of them can deal with the high dimensional quality feature set with both accuracy and easiness. In this paper, we innovatively introduce a feature selection methodology into the Critical-to-quality identification. Experimental results can verify the scientific rationality of the proposed methodology.