Simple genetic algorithm applied to the analysis of human respiratory transfer impedance data

Analysis of the human transfer impedance data (Ztr) with six-coefficient model requires using an optimisation algorithm minimising the square error criterion. Commonly used gradient algorithm do not lead to optimal solution in a single run, as the criterion function can have many local minima. So it is necessary to run the algorithm many times to be sure that the solution found is really optimal. We present an approach using single genetic algorithm applied (GA) to the analysis of Ztr measured in humans. The results from the analysis of Ztr data collected in 9 healthy volunteers in the range 2 to 64 Hz show that after a single run of GA we obtain a solution which is very close to optimal set of model parameters. As the solution found with GA is not optimal, it can be used as a starting point to classic method with gradient algorithm. We conclude, that the approach presented in the paper results in shortening of the time needed to calculate optimal parameters of the model and increases the certainty of finding optimal solution