I have never used ICA, but according to Wikipedia it looks like yet another heuristic random walking algorithm. So, it is up to you how many times to perform each kind of random shifts on each step. Moreover, it is a question of definition, what is a "step"? One trial to shift one "country"? A set of trials that shift all "countries"?

Anyway there is no much value in theoretical estimations of a single step complexity. Theoretical convergence rate estimations for random walking algorithms are usually very conservative. Especially compared to experimental results, in case of fortunate heuristics.

you can follow Mr. Atashpaz's works or papers:

Esmaeil Atashpaz Gargari

Department of Electrical and Computer Engineering Texas A&M University, College Station, TX, USA

Pattern Recognition, Proteomics, Computational Biology, Next Generation Sequencing, Optimization

tamu.edu

I checked his work but unfortunately I could not find any paper which is related to computational complexity calculation.

He is the inventor of this algorithm and he has introduced this algorithm at the first time..

I know him and I also aware that he is inventor of this algorithm but I could not find appropriate resource about my problem.