基于神经网络对葡萄风味的研究

Research on Grape Flavor Based on Neural Network

摘要: 葡萄的风味由口味和香味组成，涉及可溶性固形物等多个理化性质。在本文中，作者从6个因素入手，以经过训练的品评人员打分为结果，将风味评级划分为5个等级，提出了融合注意力机制的深度神经网络风味评价模型。模型经过训练后，预测分数与实际分数的平均差值为3.0分，等级预测准确率达到92.6%。本文在一定程度上对葡萄的评价起到了积极作用，为今后葡萄选种提供理论依据。

Abstract: The flavor of grapes is composed of taste and aroma, involving multiple physical and chemical properties such as soluble solids. In this paper, the author starts from 6 factors, divides the flavor rating into 5 grades based on the results of the trained tasters, and proposes a deep neural network flavor evaluation model integrating attention mechanism. After the model is trained, the average difference between the predicted score and the actual score is 3.0 points, and the level prediction accuracy rate reaches 92.6%. This paper has played a positive role in the evaluation of grapes to a certain extent, and provided a theoretical basis for the selection of grapes in the future.