Design of an Intelligent Recommendation System Based on Social Media Sentiment Analysis and Deep Reinforcement Learning to Improve Recommendation Quality in E-Commerce
Design of an Intelligent Recommendation System Based on Social Media Sentiment Analysis and Deep Reinforcement Learning to Improve Recommendation Quality in E-Commerce
Mohammad Mahdi Merci1 Mahmood Deypir2
1) Department of Information Technology Engineering, ST.C., Islamic Azad University, Tehran, Iran
2) Department of Computer Engineering, ST.C., Islamic Azad University, Tehran, Iran
IOI Code
:
AAHD-ABHAKD
Publication :
National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
Recommendation systems have become a key component of e-commerce platforms, playing an important role in personalizing services, increasing user satisfaction, and improving purchase conversion rates. However, many conventional recommendation methods rely solely on users’ interaction histories or item features and are unable to leverage the rich and dynamic information available on social media platforms. In addition, most of these approaches lack adaptive learning capabilities in response to continuous changes in user preferences. In this study, an intelligent recommendation system is proposed based on the integration of social media sentiment analysis and deep reinforcement learning. In the first stage, data collected from user reviews on social media and behavioral data from the Amazon Reviews dataset were gathered and preprocessed. Then, using natural language processing models, user sentiments were extracted and utilized as an influential feature in the recommendation process. Next, a deep reinforcement learning agent was designed to learn an optimal recommendation policy capable of dynamically improving personalized suggestions based on user feedback. The performance of the proposed method was evaluated using Precision, Recall, F1-score, NDCG, and conversion rate, and compared with several baseline methods. Experimental results demonstrate that the proposed model outperforms comparative approaches across all evaluation metrics. It not only improves recommendation accuracy but also enhances user engagement and increases the likelihood of purchase. The findings indicate that integrating sentiment analysis with deep reinforcement learning provides an effective framework for developing the next generation of intelligent recommendation systems in e-commerce.