系统工程理论与实践
繫統工程理論與實踐
계통공정이론여실천
Systems Engineering—Theory & Practice
2015年
1期
124~129
,共null页
李倩倩 李瑛 顾基发 刘怡君
李倩倩 李瑛 顧基髮 劉怡君
리천천 리영 고기발 류이군
舆论 级联 社会影响 可变聚类系数网络 三元无标度网络
輿論 級聯 社會影響 可變聚類繫數網絡 三元無標度網絡
여론 급련 사회영향 가변취류계수망락 삼원무표도망락
public opinion; cascading; social impact theory; tunable clustering network; triad scale freenetwork
舆论建模跳出传统的基于最近邻(“person—person”)的交互范式,引入次近邻(“person—person—person”)的影响,刻画网络中邻居的邻居对观点改变的作用,提出舆论演化的社会影响级联模型,分析其在可变聚类系数网络上舆论的演化性质.通过调节网络聚类系数,使用异步更新的方式,观察网络集聚特性对舆论演化的影响.结果表明,1)相比于传统的最近邻影响模型,社会影响级联模型的社会强化作用更大,系统更容易达成共识,初始状态中主流观点的影响将被放大;2)舆论演化结果与网络集聚性和初始状态相关:当系统初始状态P+≠P-,系统观点演化达到稳态后.网络聚类系数越大,越容易产生主流观点;当初始观点P+=P-时,即正、负力量势均力敌时,系统共识则难以确定.这种情况和现实社会舆论的演化结果符合.
輿論建模跳齣傳統的基于最近鄰(“person—person”)的交互範式,引入次近鄰(“person—person—person”)的影響,刻畫網絡中鄰居的鄰居對觀點改變的作用,提齣輿論縯化的社會影響級聯模型,分析其在可變聚類繫數網絡上輿論的縯化性質.通過調節網絡聚類繫數,使用異步更新的方式,觀察網絡集聚特性對輿論縯化的影響.結果錶明,1)相比于傳統的最近鄰影響模型,社會影響級聯模型的社會彊化作用更大,繫統更容易達成共識,初始狀態中主流觀點的影響將被放大;2)輿論縯化結果與網絡集聚性和初始狀態相關:噹繫統初始狀態P+≠P-,繫統觀點縯化達到穩態後.網絡聚類繫數越大,越容易產生主流觀點;噹初始觀點P+=P-時,即正、負力量勢均力敵時,繫統共識則難以確定.這種情況和現實社會輿論的縯化結果符閤.
여론건모도출전통적기우최근린(“person—person”)적교호범식,인입차근린(“person—person—person”)적영향,각화망락중린거적린거대관점개변적작용,제출여론연화적사회영향급련모형,분석기재가변취류계수망락상여론적연화성질.통과조절망락취류계수,사용이보경신적방식,관찰망락집취특성대여론연화적영향.결과표명,1)상비우전통적최근린영향모형,사회영향급련모형적사회강화작용경대,계통경용역체성공식,초시상태중주류관점적영향장피방대;2)여론연화결과여망락집취성화초시상태상관:당계통초시상태P+≠P-,계통관점연화체도은태후.망락취류계수월대,월용역산생주류관점;당초시관점P+=P-시,즉정、부역량세균력활시,계통공식칙난이학정.저충정황화현실사회여론적연화결과부합.
Modeling of public opinion is no longer confined to the nearest neighbourship model ('person- person' effect), we introduced the influence of ne~t nearest neighbours ('person-person-person' effect). We proposed an opinion evolution model of social cascading impact, characterized the effect of neighbours' neighbours and investigated the opinion dynamics on tunable clustering network. By adjusting the triad formation parameter which was used to change clustering coefficient of network, we applied asynchronous updating mechanism to observe the clustering coefficient influence on opinion formation. Simulation results show that: 1) compared with traditional nearest neighbour impact model, the social cascading impact model has stronger social reinforcement and as a result, is easier to reach a dominant consensus. 2) when p+ ≠ p -in initial state, a large clustering coefficient favors development of a consensus; when p+ = p- in the initial state, consensus results present uncertainty.