Multi-source data fusion strategy for the discrimination of Succus Bambusae oral liquid from different manufacturers

Abstract

Abstract This study aims at developing an extensive strategy for distinguishing Succus Bambusae oral liquid (SBOL) from different manufacturers. First, a combination of HS-GC-IMS and Ultra-fast GC E-nose method was established to perform a qualitative analysis of volatile compounds in SBOL. Second, in combination with the results of previous GC-MS studies, a multi-source data fusion strategy based on three signal sources distinguishes SBOL samples from various manufacturers. Multi-level data fusion strategies, including low-level data fusion, mid-level data fusion, and high-level data fusion which were evaluated and compared revealing their advantages and disadvantages in the classification context. The results suggest that the mid-level-SV data fusion strategy exhibits superior discrimination performance, thereby being chosen as the best data fusion strategy. Overall, this study provides a more comprehensive characterization of chemical information of SBOL samples, facilitating the improvement of SBOL quality standards and offering scientific insights and a foundation for the regulation of the SBOL market.