Abstract: 【Purpose/significance】Under the background of digital economy, intelligent recommendation algorithms are everywhere. It is increasingly obvious that it will ignite public opinion in the form of high attention and high forwarding once public opinion reversal event is reported. In order to effectively control network public opinion and avoid the risk of public opinion inversion, it is essential to investigate the evolution law of public opinion inversion considering intelligent recommendation algorithm. 【Method/process】There are six influencing factors of public opinion reversal determined, firstly, by this paper from the perspective of information ecology. Then, to define the neighbor rules of netizens with or without intelligent recommendation algorithm, the strength of social relationships and the threshold of interest difference are introduced. And evolution models of public opinion inversion with or without intelligent rec⁃ ommendation algorithm are also constructed. Finally, a set of effective simulation experiments is made in this work, which shows the ef⁃ fects of event topic sensitivity, reversal information credibility and netizens' reflective consciousness on the evolution of public opinion reversal.【 Result/conclusion】Intelligent recommendation algorithm will strengthen the influence of the primary effect on public opin⁃ ion, enhance the influence of inversion information, and weaken the inhibitory effect of netizens' reflective consciousness on public opinion polarization.【 Innovation/limitation】In this paper, an evolution model of public opinion inversion with intelligent recommenda⁃ tion algorithm is constructed based on the interest difference of Internet users. The influence of algorithm on the evolution of public opinion inversion can be further studied from the algorithm recommendation rules.