| 摘要: |
| 针对金鲳鱼(Trachinotus ovatus)贮藏过程中品质变化难以预测的问题，测定金鲳鱼片在0、3、6、9、12 ℃贮藏条件下挥发性盐基氮质量分数(w(TVB-N))、菌落总数、K值和感官评价值，构建径向基函数神经网络(Radial basis function neural network，RBFNN)和反向传播神经网络(Back propagation neural network，BPNN)预测模型以预测品质，并对模型的预测结果进行残差分析和相对误差分析以评价预测准确度。结果表明:1)BPNN模型和RBFNN模型的残差都是随机且不规则的，说明2种模型都适用于预测金鲳鱼片的新鲜度，但RBFNN模型残差绝对值更小；2)对于4 ℃贮藏条件下金鲳鱼片的各项品质指标，BPNN模型预测相对误差绝对值小于15%(除K值第0天)，RBFNN模型预测相对误差绝对值大部分小于5%，RBFNN模型预测相对误差绝对值较小。对于金鲳鱼片新鲜度的预测，RBFNN模型准确度较高，BPNN模型准确度较低，RNFNN模型更适合用于预测金鲳鱼贮藏品质。 |
| 关键词: 金鲳鱼 贮藏 品质变化 神经网络预测模型 |
| DOI：10.11841/j.issn.1007-4333.2023.03.12 |
| 投稿时间：2022-06-25 |
| 基金项目:海南省重点研发计划(ZDYF2021XDNY154)；国家重点研发计划(2018YFD0901001) |
|
| Construction and evaluation of BPNN and RBFNN prediction models of golden pompano fillets during storage |
|
ZHANG Jingxuan1,LIANG Shijie1,WU Xinning1,Li Mingze1,LUO Yongkang1,2*
|
| (1.College of Food Science and Nutritional Engineering, China Agricultural University, Beijing 100083, China;2.Sanya Institute, China Agricultural University, Sanya 572000, China) |
| Abstract: |
| To solve the problem that the quality change of golden pompano(Trachinotus ovatus)is difficult to predict during storage, TVB-N mass fraction(w(TVB-N)), total viable counts, K value and sensory assessment were investigated and Radial Basis Function Neural Network(RBFNN)and Back Propagation Neural Network(BPNN)prediction models were established. Residual analysis and relative error analysis were taken to evaluate and compare the accuracy of the models. The results showed that: 1)The residuals of BPNN model and RBFNN model were random and irregular, indicating that both models were suitable for predicting the freshness of golden pompano fillets, the absolute value of residuals of RBFNN model was smaller; 2)For the quality indicators of golden pompano fillets stored at 4 ℃, the absolute value of relative error predicted by BPNN model was less than 15%(except for the 0th day of K value), and the absolute value of relative error predicted by RBFNN model was mostly less than 5%, the absolute value of relative error predicted by RBFNN was smaller. For the prediction of the freshness of golden pompano fillets, the RBFNN model was more accurate, while the BPNN model was less accurate. Therefore, the RNFNN model is more suitable for predicting the storage quality of golden pompano. |
| Key words: golden pompano storage quality change neural network prediction model |
