为使草莓感官评价结果更客观，不受评价者评价经验、年龄及健康状况等主观因素影响，该研究将草莓的十项理化风味指标如过氧化氢酶、多酚氧化酶等作为输入数据，感官评价得分作为输出数据，运用灰狼优化支持向量机建立草莓感官评价估计模型。为验证所提出模型的优越性，将其与粒子群优化支持向量机模型、卷积神经网络、长短时记忆网络模型进行对比分析，为充分保证所提出模型的有效性，重复实验20次并计算各项指标均值，得到所提出模型的均方根误差和平均绝对误差分别为0.28、0.24（低于粒子群优化支持向量机的误差指标0.46、0.38，长短时记忆网络误差指标1.24、0.99和卷积神经网络误差指标0.88、0.75）。实验结果表明，基于灰狼优化支持向量机模型的预测精度最高、稳定性最好、误差最小，研究结果可为草莓感官评价得分的评定提供参考。

In order to make the result of sensory evaluation of strawberry more objective and independent of subjective factors such as evaluator's evaluation experience, age and health status, in this study, ten physicochemical flavor indicators of strawberry, such as catalase and polyphenol oxidase, were used as the input data, and the sensory evaluation score was used as the output data. The grey wolf optimizer-based support vector machine was used to establish a sensory evaluation prediction model of strawberry. To verify the superiority of the proposed model, this established model was compared with the particle swarm optimization support vector machine model, convolutional neural network model and long-/short-term memory network model. In order to ensure fully the effectiveness of the proposed model, the experiment was repeated 20 times and the mean value of each index was calculated. The root mean square error and the mean absolute error of the proposed model were 0.28 and 0.24, respectively (lower than the error indices for particle swarm optimization support vector machine (0.46 and 0.38), for long short time memory network (1.24 and 0.99), and for convolutional neural network (0.88 and 0.75). The experimental results showed that the grey wolf optimizer-based support vector machine model had the highest prediction accuracy, the highest stability, and the smallest error. The results can provide a reference for the evaluation of strawberry’s sensory evaluation score.

国家重点研发计划项目（2022YFF1101103）；北京市教育委员会科技研究计划项目资助（KM202210011006）；北京市自然科学基金青年科学基金项目（6204036）；国家自然科学基金项目（31972191）