Some of the challenges that plague the interior design sector include the inability of the users to forecast the impact of decoration, low matching ratio of design schemes to individual preferences, and time-consuming decision making. Due to these problems, this paper creates a 3D virtualized recommendation model founded on the Collaborative Filtering (CF) algorithm. First, the preference data of interior design is gathered on 1500 users to come up with a user-design scheme rating matrix. The user similarity is calculated by Pearson correlation coefficient, where the first K neighbor users are referred to as the most similar ones. Second, a 3D visualization engine is constructed based on the WebGL technology that can be used to convert suggested design schemes into interactive virtual spaces that allow real-time roaming and material manipulation. Thereafter, the interest of the target user in unrated schemes is predicted with a weighted average technique and the Top-N schemes are selected to be rendered and displayed in 3D. Lastly, item collaborative filtering is presented as a hybrid recommendation approach to enhance the accuracy of a recommendation during cold-start conditions. The experimental findings indicate that the system has a recommendation accuracy of 87.3 percent, a mean absolute error (MAE) of 0.62, stable 3D scene rendering frame rate of 66.9 fps, and a total interaction response latency of 72 ms, which is effective in enhancing the quality of customized recommendations and user experience.
Sheng et al. (Thu,) studied this question.