An anomaly detection approach based on hybrid differential evolution and K-means clustering in crowd intelligence

Abstract

Purpose – Artificial intelligence is gradually penetrating into human society. In the network era, the interaction between human and artificial intelligence, even between artificial intelligence, becomes more and more complex. Therefore, it is necessary to describe and intervene the evolution of crowd intelligence network dynamically. This paper aims to detect the abnormal agents at the early stage of intelligent evolution. Design/methodology/approach – In this paper, differential evolution (DE) and K-means clustering are used to detect the crowd intelligence with abnormal evolutionary trend. Findings – This study abstracts the evolution process of crowd intelligence into the solution process of DE and use K-means clustering to identify individuals who are not conducive to evolution in the early stage of intelligent evolution. Practical implications – Experiments show that the method we proposed are able to find out individual intelligence without evolutionary trend as early as possible, even in the complex crowd intelligent interactive environment of practical application. As a result, it can avoid the waste of time and computing resources. Originality/value – In this paper, DE and K-means clustering are combined to analyze the evolution of crowd intelligent interaction.