Title: Research on big data personalised recommendation model based on deep reinforcement learning

Authors: Haifeng Shi; Ling Shang

Addresses: School of Network and Communication, Nanjing Vocational College of Information Technology, Nanjing, 210023, China ' School of Network and Communication, Nanjing Vocational College of Information Technology, Nanjing, 210023, China

Abstract: In order to mine the user's preference and interest from the user's historical behaviour in the big data to make a personalised recommendation, a DRR model is constructed based on deep reinforcement learning, and the performance of the DRR model is analysed through experiments. The results showed that the DRR model had a higher effect than other comparable models in the offline experimental evaluation, and the DRR-att value was the highest, reaching 0.9025. In the online simulation experiment, the average DRR-att value was the highest reward rate, reaching 0.7466. In general, the DRR model had better analysis ability and strong dynamic modelling ability and was good at using long-term rewards for decision making. In the parameter analysis experiment, the T value reached ten points. At the same time, the user state expression module can improve the accuracy of the DRR model and is effective in actual user personalised recommendations.

Keywords: deep reinforcement learning; personalised recommendation; dynamic modelling; effectiveness.

DOI: 10.1504/IJNVO.2023.133876

International Journal of Networking and Virtual Organisations, 2023 Vol.28 No.2/3/4, pp.364 - 380

Received: 21 Dec 2022
Accepted: 12 Jun 2023
Published online: 04 Oct 2023 *

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