This study researches the influence of two critical group characteristics, i.e., knowledge distribution and group structure, on the effect of collective intelligence emergence (group performance). We introduce a group structural measure based on a group collaboration network and a knowledge distribution index based on decision variables. Group structure mainly includes link intensity and the degree to which the network is hierarchical or flat. The knowledge distribution status classifies groups into three types: generalist groups, mixed groups and specialist groups. We adopt a continuous-time Markov chain based on NK fitness landscape theory to model the ability of groups to solve complex system decision-making problems. Group members revise their opinions driven by two factors: individual interest(i.e., the pursuit of perceived payoff improvement) and social influence(i.e., the pursuit of consensus seeking). The results show that: 1) Unbalanced knowledge distribution will undermine group performance. 2) However, group structure will regulate the influence of knowledge distribution. 3) Generalist groups have a better performance than mixed groups and specialist groups when addressing complex problems. 4) It is not suitable for groups to seek a very high consensus level.