Model for estimating the weight-loss ratio of damaged Korla fragrant pears

Abstract

To predict the weight-loss ratio of Korla fragrant pears effectively, improve commodity value and study the variation laws of the weight-loss ratio of damaged fragrant pears during storage, this study predicted the weight-loss ratio of fragrant pears by utilizing the generalized regression neural network (GRNN), support vector regression (SVR), partial least squares regression (PLSR) and error back propagation neural network (BPNN). The prediction performances of GRNN, SVR, PLSR and BPNN models were compared comprehensively, and the optimal model was determined. In addition, the optimal prediction model was verified. The results show that weight-loss ratio of fragrant pears increases gradually with the extension of storage time. During storage, the weight-loss ratio of fragrant pears is positively related to the degree of damage. The trained GRNN, SVR, PLSR and BPNN models can be used to predict the weight-loss ratio of fragrant pears. The BPNN model is the most accurate in predicting the weight-loss ratio of damaged fragrant pears (R 2 =0.9929; RMSE=0.2138). It has also been proved to have good predictive effect in production practice (R 2 =0.9377, RMSE=0.7138). The research findings can provide references to predict the delivery quality and time of delivery of Korla fragrant pears.