Article: The application of K-means based user feature clustering algorithm in graphic design interface Journal: International Journal of Web Engineering and Technology (IJWET) 2025 Vol.20 No.4 pp.404 - 421 Abstract: The graphic design interface needs to provide users with a clear and appropriate interface based on their behaviour, which requires the platform to achieve personalised recommendation. To address this issue, this study proposes a collaborative filtering algorithm based on K-means user feature clustering. Firstly, an overall framework is designed based on collaborative filtering, introducing K-means clustering algorithm to optimise the initial model cold start and data sparsity issues. Finally, in order to respond more accurately when user preferences change, the concept of time window is introduced, and the user project matrix is re-evaluated by time weighting.. The experiment indicated that the accuracy of the personalised recommendation model proposed in the study reached 97.82%. This was an average improvement of 7% compared to type-based classification model, deep neural network based on logistic regression and feedforward, and semantic relationship temporal recommendation based on deep collaborative filtering. The average recall rate increased by 10.09%, the F1 value increased by 8.62%, and the coverage increased by 6.4%. In summary, the collaborative filtering algorithm on the basis of K-means user feature clustering can achieve a clear and personalised precise graphic design interface. Inderscience Publishers - linking academia, business and industry through research

Title: The application of K-means based user feature clustering algorithm in graphic design interface

Authors: Xiaohui Jia; Yiyi Zhao

Addresses: Department of Art and Design, Shijiazhuang College of Applied Technology, Shijiazhuang, 050800, China ' Department of Art and Design, Shijiazhuang College of Applied Technology, Shijiazhuang, 050800, China

Abstract: The graphic design interface needs to provide users with a clear and appropriate interface based on their behaviour, which requires the platform to achieve personalised recommendation. To address this issue, this study proposes a collaborative filtering algorithm based on K-means user feature clustering. Firstly, an overall framework is designed based on collaborative filtering, introducing K-means clustering algorithm to optimise the initial model cold start and data sparsity issues. Finally, in order to respond more accurately when user preferences change, the concept of time window is introduced, and the user project matrix is re-evaluated by time weighting.. The experiment indicated that the accuracy of the personalised recommendation model proposed in the study reached 97.82%. This was an average improvement of 7% compared to type-based classification model, deep neural network based on logistic regression and feedforward, and semantic relationship temporal recommendation based on deep collaborative filtering. The average recall rate increased by 10.09%, the F1 value increased by 8.62%, and the coverage increased by 6.4%. In summary, the collaborative filtering algorithm on the basis of K-means user feature clustering can achieve a clear and personalised precise graphic design interface.

Keywords: K-means clustering; user attributes; interest preferences; time; collaborative filtering; CF; graphic design.

DOI: 10.1504/IJWET.2025.151163

International Journal of Web Engineering and Technology, 2025 Vol.20 No.4, pp.404 - 421

Received: 15 Apr 2024
Accepted: 26 Jan 2025
Published online: 15 Jan 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: The application of K-means based user feature clustering algorithm in graphic design interface

Keep up-to-date