In order to meet consumer demand for ceramic bottle design, the mapping relationship between user imagery needs and ceramic bottle product form design elements is established by analysing them. Firstly, the representative samples are screened by the Delphi method, and the stylistic features of ceramic bottles are summarised by using the principle of visual simplification, secondly, the elements of ceramic bottles are modularly deconstructed by using the morphological analysis method, and the importance of the product elements is calculated by the combination of the DEMATEL and the TODIM method, and the evaluative value of the pairs of imagery words is obtained by using the Likert scale method, and representative imagery words are determined by using the KJ method and the factor analysis method. Vocabulary. A BP neural network was used to construct the modelling imagery prediction model, and the model was trained with the weights of the sample modelling elements as the input layer, and the evaluation scores of the user’s perceptual imagery as the output layer. According to the training and validation of the model, the optimal ceramic bottle shape with the best perceptual evaluation value can be predicted. This method solves the problem of designers not being able to grasp the needs of target users, improves the design efficiency of ceramic bottle products, and provides favourable support for the personalised design of ceramic bottle products.