نوع مقاله : مقاله پژوهشی
نویسندگان
1 کارشناسی ارشد ، دانشگاه جامع امام حسین(ع)، تهران، ایران
2 استاد ، دانشگاه جامع امام حسین(ع)، تهران، ایران
3 دانشجوی دکتری ، دانشگاه جامع امام حسین(ع) ، تهران ، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
The analysis of online social network data represents a significant scientific challenge in contemporary research. Within the domain of user opinion analysis in these networks, dynamic opinion maximization has emerged as a nascent field of study. Prior investigations in this area have predominantly operated under the assumption of static, unchanging node opinions. Furthermore, the temporal evolution of user perspectives has received comparatively limited attention. This paper proposes a method for the dynamic opinion maximization problem, based on a greedy genetic algorithm, explicitly considering the dynamics of opinions and their evolution over time.The proposed method comprises two principal components: an activated opinion dynamics model and a seed node selection process. The activated opinion dynamics model is constructed by integrating the linear threshold model with stateless Q-learning, thereby explicitly accommodating the temporal fluctuations in opinions. A greedy genetic algorithm is employed for the selection of seed nodes. Following the identification of an initial seed node, the activated opinion dynamics model is initiated. During this phase, the seed node and its immediate neighbors are activated according to the linear threshold model. Subsequently, stateless Q-learning is utilized to update the opinions of the nodes. This iterative process continues until predefined termination criteria are satisfied. Experimental results on four signed social network datasets demonstrate that the proposed framework outperforms the state-of-the-art method by 14% in terms of the number of activated nodes and 27% in terms of average positive opinions.
کلیدواژهها [English]
)