Grade classification model tandem BpNN method with multi-metal sensor for rice eating quality evaluation

•A novel tandem evaluation method was developed for E-tongue to predict rice score.•Tandem evaluation method consisted of grade classification model and BpNN.•The performance of tandem evaluation method was better than direct neural network.•The result obtained with the method was consistent with that from sensory evaluation. Evaluation of rice eating quality is increasingly important to researchers and consumers. In this paper, a novel tandem evaluation approach, which is grade classification model tandem back propagationartion neural network (BpNN) method, was developed for multi-metal sensor to predict rice eating quality. Characteristic current and potential arrays were extracted and properly processed, resulting in potential phasor plane values and current phasor plane values. The current phasor plane values were used as the input variables of grade classification model, and consequently, the classification value and grade value of sample output as well as potential phasor plane values were used in the following BpNN for score prediction. It was found that the performance of developed grade classification model tandem BpNN method was better than that of direct neural network method. The scores predicted by developed method were compared to sensory scores obtained by traditional sensory evaluation. A correlation coefficient of 0.9964, a RMSE of 0.69, a slop of 0.98 and an intercept of 1.45 were obtained. The accuracy of developed method for score prediction was 90%. Results indicated that the developed tandem evaluation approach was an efficient method for rice score prediction in electronic tongue system.