¹School of Information Engineering of Wuhan University of Technology, Wuhan, PR. China.
²Aviation Industry Corporation of China, Beijing, PR. China.

---

Abstract—Feature extraction or feature selection is the premise and key of rule mining and fault diagnosis in fuzzy information system. According to the characteristics of large-scale fuzzy information system, such as dynamic, large amount of data, fuzziness and high dimensionality, this paper is based on dynamic reduction method, used dynamic sampling method, divided the large data set into small data set, and transformed the dynamic information system into a series of static information system. At the same time, this paper is used an intuitionistic fuzzy-rough set method, proposed a generalized dynamic feature extraction algorithm based on the theory of discernibility matrix of intuitionistic fuzzy-rough sets and dynamic reduction. The algorithm obtains the key fault feature parameters of fuzzy decision information system with dynamic, large amount of data, fuzziness and high dimensionality. Taking the actual sampled aero-engine data sets as an example, the algorithm is proved to be scientific, effective and correct. Under the condition of guaranteeing the accuracy of diagnosis, the minimum attribute set obtained by the algorithm is proved to be the characteristic parameter of the fuzzy decision information system, and the size of rule base can be reduced by 99.2%. The algorithm can be used for aircraft fault classification, fault diagnosis and condition monitoring in big data environment.

Index Terms—Intuitionistic fuzzy-rough set, discernibility matrix, generalized dynamic reduction, feature selection, big data.

[PDF]

Cite: Zhang Chuanchao, "Large Data Generalized Dynamic Fault Feature Extraction Algorithm Based on Intuitionistic Fuzzy-Rough Set Discernibility Matrix," Journal of Computers vol. 14, no. 1, pp. 1-24, 2019.