Research On The Problem Of Influence Maximization Across Networks

Influence maximization is a seed-nodes selection problem based on viral marketing and information propagation model.It is widely used in advertising,rumor warning and other fields,and has become a research hotspot in social network analysis.However,most of the existing research usually focus on a specific social network,which means that they only analyse various types of data in only one network.This can result in low accuracy of the evaluation of user influence and insufficient influence spread.For this reason,this thesis takes the research of influence maximization in a single network as the starting point,and studies the problem of influence maximization across networks.Firstly,the problem of influence maximization across networks is based on crossnetwork propagation of influence,which requires cross-network users among different networks as the intermediary,so the first research goal of this thesis is to identify crossnetwork users among different networks efficiently and accurately.Secondly,on the premise of ensuring good time efficiency,selecting more efficient seed-nodes in crossnetwork environment than in single-network environment so as to solve the problem of influence maximization across networks is another goal of this thesis.Specifically,the research contents of this thesis are as follows:(1)Aiming at the inefficiency of existing algorithms in matching the whole dataset by pairwise and the low accuracy caused by ignoring the characteristics of user' interests while identifying cross-network users,an identity recognition algorithm for cross-network users based on users' interests(UI-UI)is studied and designed.Firstly,the Blocking idea was used to perform preliminary screening on users' accounts to improve the efficiency of the algorithm.Secondly,in the process of identifying the identities of cross-network users,both topic interest and interactive interest of users were taken into account as the recognition basis,and the negative impact of false matching on the subsequent recognition results was reduced by a process called reverse verification,so as to improve the accuracy.Finally,the experimental results show that UI-UI algorithm can effectively identify cross-network users,and its time efficiency,accuracy and recall rate are better than the benchmark model.(2)Aiming at the problem of insufficient influence spread and low efficiency of influence propagation in single-network environment,a Two-Phase Algorithm for Influence Maximizing across Network(TP-IMaN)is studied and designed.Firstly,an Interest-Driven Influence Propagation Model across networks(ID-IPM)was proposed.In order to improve the range of influence propagation,this model not only coupled networks according to the "bridge" role of the cross-network users,but also calculated the influence strength between users by considering both the particularity of information propagation within and between networks and the differences of users' interests.Secondly,based on ID-IPM model,a heuristic and greedy hybrid algorithm was designed to select seed-nodes in two stages to improve the efficiency of the algorithm.Finally,the experimental results show that the TP-IMaN algorithm in crossnetwork environment can select seed-nodes more effectively than the benchmark algorithms in single-network environment,and TP-IMaN has wilder influence spread than the benchmark models.