Abstract: Aiming at the problem of long time-consuming selection and placement of accessories in the home design process, a data-driven smart jewelry displayed method that integrates design history knowledge is proposed. Firstly, the home scene designed by the existing professional designers is divided into a series of sub-space structures, and the sub-spaces for storing the accessories are marked and the corresponding information about the characteristics of the accessories and the sub-spaces is extracted. Then the corresponding learning model is constructed, and the feature information about the subspace is combined with the feature information about the jewelry to form a fusion feature. And then, corresponding jewelry placement plan for the series subspace where jewelry needs to be placed in a new scene is generated by the trained model and the jewelry characteristics in the jewelry library. At the same time, the combination of the jewelry in the subspace is given according to the location of the jewelry in the existing display history data and the semantic information mining the relationship between the jewelry. The experimental results show that the accuracy of Top-10 and the recall rate of Top-10 are 0.166 9 and 0.432 0, respectively. Compared with the neural collaborative filtering model, the proposed method is improved, and it can give a reasonable jewelry placement recommendation for indoor home which can greatly save the user’s time for placing jewelry.