Sensory Evaluation Based on Ensemble Learning

Sensory evaluation is one of the key steps in recipe product design. With the development of compute intelligence technology, many methods such as artificial neural network, decision tree, regression, etc are used to solve the problems in sensory evaluation. This becomes more and more popular. But the generalization ability using single model needs to be improved. This paper uses bagging algorithm for ensemble learning to carry out sensory evaluation and compares it with single classifier m5p. Through feature selection, we improve the accuracy and decrease the complexity of ensemble learning and find a compromise structure. The results of experiment prove that ensemble learning can improve the generalization ability, that is to say, ensemble learning is superior to single classifier in sensory evaluation.