Title: Research on the construction of personalised recommendation system based on LFM algorithm and tag data

Authors: Guofang Liu; Lixiang Shi

Addresses: School of Computer Science, Sichuan University Jinjiang College, Meishan City, Sichuan Province, 620860, China ' School of Computer Engineering, Chengdu Technological University, Chengdu City, Sichuan Province, 611730, China

Abstract: In view of the low efficiency of the current personalised recommendation system, this research proposes a personalised recommendation method combining latent factor model (LFM) algorithm and label data. In this recommendation method, LFM algorithm decomposes the user click matrix by using the gradient descent method to complete the prediction of implicit feature feedback. The tag data can be predicted by explicit feature feedback through BM25 algorithm, item tag matrix and user scoring matrix. The results show that the performance of LFM algorithm is the best when the implicit feature dimension is 30. At this time, the accuracy, recall, coverage and F1 values were 27.82%, 8.82%, 53.32% and 19.86% respectively. The variation range of accuracy is 23.06~26.67%, and the variation range of recall is 8.42~10.03%. The application of the research results in the personalised recommendation system can obtain a more efficient and comprehensive recommendation technology.

Keywords: personalised algorithm; LEM; latent factor model; interest characteristics; BM24; tag data.

DOI: 10.1504/IJCSM.2023.135049

International Journal of Computing Science and Mathematics, 2023 Vol.18 No.4, pp.350 - 364

Received: 13 May 2022
Accepted: 11 Apr 2023
Published online: 28 Nov 2023 *

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