Community Opinion Network Maximization for Mining Top K Seed Social Network Users
Lecture Notes in Networks and Systems
Community Detection, Opinion Maximization, Opinion Mining, Sentiment Classification, Social Network, Viral Marketing
The Active Community Opinion Network Mining and Maximization (ACOMax) system being proposed improves on the capabilities of an existing system, Opinion Based Influence Network (OBIN) system by integrating active opinion mining (OM) as in the system, aCtive OpinioN Estimator (CONE). This work also adds opinion maximization to enhance OBIN’s accuracy while using joint aCtive OpinioN Estimator (CONE’s) training time to solve OBIN’s cold-start problem. It does this using joint sentence-level and feature-level opinion mining of mined community opinions for filtering out non-positive opinions. ACOMax first mines multiple posts related to a product using Twitter API, before performing joint opinion mining on selected posts from these user reviews. From the selected posts, ACOMax obtains frequent features with favorable opinions of the product, which it uses to construct a community opinion network graph of users sharing positive opinions. The graph is used by the seller to actively find the top k seed users to produce maximum opinion spread when Multiple Linear Threshold (MLT) is used for opinion maximization. The proposed system improves total opinion spread in a social network over the compared systems. When a small top k = 10 seed users is used, the experimental results show a total opinion spread of 263 users for ACOMax system in comparison to the 112 users produced by the system CONE for a 134 % improvement in opinion spread. ACOMax also improves on OBIN system’s accuracy as while it achieves respective higher scores of 98.19 %, 98.50 %, and 97.89 % for F1, precision and recall; OBIN system achieves 95.3 %, 98.24 %, and 93.71 %.
Ezeife, Christie I. and Semwal, Mayank. (2023). Community Opinion Network Maximization for Mining Top K Seed Social Network Users. Lecture Notes in Networks and Systems, 700 LNNS, 3-16.