Prediction Model of the Main Chemical Components from the Upper Tobacco Leaf Curing Process Based on Dendritic Network

Purpose To monitor the content of main chemical components in tobacco leaves during curing process in real time.

Methods The prediction model of main chemical components content based on dendritic network was constructed by the color value of tobacco leaves during curing process, and the model was compared with BP neural network and stepwise regression.

Results The model based on dendritic network performed well in both the training set and the validation set, the root mean square error (RMSE) and mean absolute percentage error (MAPE) values were lower, and the coefficient of determination (R²) was close to one, indicating that the model had good fitting degree and stability, and it was better than BP neural network and stepwise regression. The model performs well in external prediction, indicating that the model had higher fitting degree and better prediction performance.

Conclusion The dendritic network can be used for real-time monitoring of the main chemical components in the tobacco curing process, timely regulating the curing process, improving the quality of cured tobacco leaves, and providing reference for future intelligent tobacco curing.