该研究运用主成分分析(PCA)结合误差反向传播(BP)神经网络对啤酒感官评价进行了预测。把啤酒中11种理化及风味指标进行主成分分析,以除去数据之间的线性相关性,提取后的理化及风味指标做为输入数据,感官评价得分作为输出数据,运用BP神经网络建立啤酒感官评价预测的模型。使用此模型对50种啤酒的感官得分进行预测,预测最大相对误差为2.68%。结果表明,主成分分析和神经网络相结合的这种方法能够准确预测啤酒感官评价得分。

The method integrating principal component analysis(PCA) with back propagation(BP) neural networks was applied to predict the beer sensory evaluation.A total of 11 physicochemical indexes was extracted by PCA to eliminate linear relevance and then used as input.Sensory evaluation scores were used as output.BP neural network was established to model the beer sensory evaluation.Finally,sensory evaluation scores of 50 beers were predicted by this model.The experimental results showed that the maximum predictive relative error was 2.68%,suggesting that the methods integrating PCA with BP networks could precisely predict the beer sensory evaluation.