一种新型社会网络影响最大化算法_田家堂.pdfVIP

34 10 V ol . 34 N o. 10 20 11 10 CH IN ESE JO U RN A L O F CO M PU T ER S Oct . 2011 田家堂 王轶彤 冯小军 ( 20 1203) k , k . N P- . Kemple Kleinber g , , , . , , H PG. H PG , PI , . H P G . T P 311 DOI : 10. 37 24/ SP. J. 10 16 . 2011. 0 1956 A New Hybrid Algorithm for Influence Maximization in Social Networks T IA N Jia- T ang WA N G Y-i T ong FEN G X iao-Jun ( S chool of Comp u ter S ci ence , Fud an Univ er s ity , Sh anghai 20 1203) Abstract Infl uence max imizat io n is a problem of finding a sm all subset o f nodes ( t ar get set) in a social net w ork th at could max imize t he spread of inf lu ence. T his opt im ization problem of inf lu- ence maxim iz at ion is N P- h ar d under sev er al m ost w idely stu died diff usion m odels and is even ch al- lenging fo r current online social net w ork s w hich cont ain bot h positiv e and neg at ive relation s. K em ple and Kleinberg pro posed a nat ur al climbing- hill gr eedy alg orit hm t hat choo ses t he nodes w hich could provide a go od marg inal inf lu ence. T his greedy algo rithm h as larg e spread of inf lu- ence, but is very cost ly and cannot be applied t o large so cial net w ork s. A lso, greedy alg orit hm could not gu arant ee the best influ ence spr ead. In t his pap er, w e propose a f ram ew o rk o n t he line- ar t hresho ld mo del and a hybrid po tential- inf luence greedy algo rithm ( H P G) w hich can m ake goo d u se of t he inf lu ence accumulation pro p