The declining birth rate side by side reflects the people’s demand for high-quality childcare. Guided by that demand, the effective design of community infant and toddler public spaces is particularly important. In this paper, we collect text data of social networking site comments related to the sensory experience of community infant and toddler public spaces from the target audience to construct a research dataset. Utilizing the preprocessed standard data, LDA topic model is established and combined with semantic relatedness calculation to determine the optimal number of topics. After that, through the random forest algorithm using CART decision tree as a class learner, the text data were categorized to mine the highest value design features, and these design features were used as the basis for enhancing the experience of public space for infants and toddlers in the community.The number of themes outputted from the LDA model was four. On this basis, the random forest algorithm mines the 10 best design features including visual. Based on the 10 features, we completed the design and remodeling of a number of public spaces for infants and toddlers in District Y. The final evaluation of the public spaces by the target population on all dimensions reached an average score of 4 or more. The design of public spaces for infants and toddlers based on sensory experience is effective.
