Development and industrial application of geographical origin identification for Taiwanese oolong tea.

Abstract

Taiwanese oolong tea is renowned for its excellent quality and enjoys a prestigious reputation both domestically and internationally. In recent years, there has been an issue with imported Taiwanese-style oolong tea being sold as genuine Taiwanese oolong tea, which has adversely affected the brand value of Taiwanese oolong tea. In this study, samples of domestic oolong tea (Taiwanese oolong tea) and Taiwanese-style oolong tea produced abroad (including China, Vietnam, Indonesia, Thailand, etc.) were collected. A multi-elements analysis method was applied to establish an elemental database of tea leaf samples. Subsequently, various widely used classification methods were employed to develop a discrimination model for identifying the origin of Taiwanese oolong tea. Utilizing the discrimination model established from a database of 727 samples to determine whether the tea leaves were Taiwanese or external, the statistical performances of classification models such as LDA, Ridge, Random Forest, Boosting, and SVM are nearly consistent. These models achieved an accuracy rate of 97.1%-97.8%, a recall rate (true positive rate) for Taiwanese origin of 98.4%-99.0%, and a precision value for predicting Taiwanese origin of 97.3%-97.8%. This identification technology has become an officially recognized and publicly recommended testing method in Taiwan (TFDAF0032.00, released on November 5, 2021) and has been effectively utilized in official administrative inspections for identification of origin, as well as providing evidence for investigative cases.