Int J Performability Eng ›› 2019, Vol. 15 ›› Issue (8): 2267-2276.doi: 10.23940/ijpe.19.08.p27.22672276

Previous Articles     Next Articles

Collaborative Filtering Algorithm based on Data Mixing and Filtering

Xiaohui Chenga,b, Li Fenga,b, and Qiong Gui a,b,*   

  1. a College of Information Science and Engineering, Guilin University of Technology, Guilin, 541000, China
    b Guangxi Key Laboratory of Embedded Technology and Intelligent System, Guilin, 541000, China
  • Submitted on ;
  • Contact: * E-mail address: cxiaohui@glut.edu.cn
  • About author:Xiaohui Cheng received his bachelor's degree from Shanghai University of Technology in 1982. He is currently the dean of the School of Information Science and Engineering at Guilin University of Technology. He is also the director of the Key Laboratory of Guangxi University of Information and Manufacturing, the executive director of the Guangxi Computer Society, the director of the Embedded Systems Branch of the China Computer Software Industry Association, the executive director of the Guangxi Computer Software Industry Association, a member of the Guangxi Natural Science Foundation Expert Committee, and the vice chairman of the Guangxi Measuring Instruments Industry Alliance. His current research interests include embedded systems and Internet of things technology. Li Feng is a graduate student in the School of Computer Science and Technology at Guilin University of Technology. Her research interests include embedded systems and Internet of things technology. Qiong Gui received her master's degree from Guilin University of Technology and her Ph.D. from the School of Information Engineering at Wuhan University of Technology. She is currently an associate professor and graduate instructor in the College of Information Science and Engineering. Her research interests include big data analysis and information security.
  • Supported by:
    This work is sponsored by the National Natural Science Foundation of China (No. 61662017, 61862019, and 61262075,), and Guilin Science and Technology Project Fund (No. 2016010408).;

Abstract: Personalized recommendation systems based on the collaborative filtering algorithm are faced with an excessive user rating data sparseness problem. In order to solve this problem, an improved collaborative filtering algorithm is proposed, which gathers a variety of single numerical filling methods and selects a more appropriate filling method according to the filling rules to fill the vacant positions in the user-item scoring matrix filling. The recommendations are then made on the populated user-item score matrix through a user-based collaborative filtering approach. The method of data mixed filling can effectively reduce the recommended error and numerical singularity caused by fixed filling values such as the mean and median. The improved collaborative filtering algorithm is tested on the Movie Lens data set. The results show that the method of data mixing is adopted to fill the empty positions in the scoring matrix, which effectively alleviates the data sparsity problem in the collaborative filtering algorithm and improves the accuracy of recommendation systems for target users.

Key words: collaborative filtering, sparse data, multivariate value filling, fill rule