Attribute reduction is crucial for resource-constrained microsystems to achieve efficient and accurate decision-making.However

traditional attribute reduction methods struggle to handle hybrid information systems with coexisting continuous and discrete data.To address this issue

a hybrid data attribute reduction method based on fuzzy similarity relation conditional entropy is proposed.Given the characteristics of hybrid information systems

equivalence relations and neighborhood relations are integrated and improved first

and a novel fuzzy similarity relation is proposed.Building upon this relation

a fuzzy similarity relation granular structure model is established for hybrid information systems

and the metrics for fuzzy similarity relation information quantity and information entropy are defined.Finally

by utilizing the fuzzy similarity relation conditional entropy metric model

a new attribute reduction algorithm is designed.Experimental results on typical medical sensing diagnosis datasets demonstrate the effectiveness of the proposed algorithm

which provides a practical approach for lightweight processing and intelligent analysis of sensing data in microsystems.