Selecting variables with the least correlation based on physarum network

In spectroscopy, redundant information makes the number of input variables for a prediction model larger than required. We present a method based on the physarum network to select the variable with the least correlation. This method transforms the variable selection problem into a path finding problem and then solves the problem based on the mechanism of foraging of Physarum polycephalum. Experimental results show that the physarum network, combined with other feature selection or extraction methods, can select the least number of wavelengths without sacrificing the prediction performance.