PREDICTION OF MICROBIAL INACTIVATION IN UV LIGHT TREATMENT OF WHITE TEA USING MACHINE LEARNING AND NEURAL NETWORKS.

The potential of ultra-violet (UV) light to replace the traditional brewing process to make cold tea in terms of inactivation of endogenous microflora has not been explored. Thus, the efficacy of emerging technologies such as UV-C by tea leaves/water ranging from 1 to 3 %, number of lamps ranging from 2 to 8, and number of cycles ranging from 4 to 8 were performed to determine the inactivation of total mesophilic aerobic bacteria (TMAB) and total mold and yeast (TMY) and changes in quality properties in cold drip white tea. The UV-light process was effective to reduce both TMAB and TMY. Increased number of cycles provided a significant amount of inactivation on both TMAB and TMY. The reduction of initial number of TMY was determined as 3.40±0.03 log cfu/mL with the number of lamps of 5, the number of cycle of 4, and tea leaves/water ratio of 1%, whereas TMAB were found as 3.12±0.08 log cfu/with the number of lamps of 2, the number of cycles of 6 and tea leaves/water ratio of 1%. The resulting datasets were used to predict the inactivation of TMAB and TMY in cold drip white tea using gradient boosting regression tree (GBRT), random forest regression (RFR), and artificial neuron network (ANN) models. The ANN model provided the lowest RMSE and highest R2 value for predicted inactivation of TMAB. TMY has not been predicted using either machine or neural networks. UV treatment possess a viable alternative for microbial inactivation without adverse effect on the quality properties of cold drip white tea. [ABSTRACT FROM AUTHOR]