Robust grey model based on genetic algorithms and its application to prediction for chromatographic retention

The routine modelling method for GM(1,1) grey model is described in details and its disadvantages are pointed out. The annealing evolution algorithm derived from simulated annealing and combined with the strategy of evolution in the genetic algorithm is applied to GM(1,1) modelling, and the uniform design method is adopted in the algorithm to optimize the important parameters N (the population), L (the length of Markov chain) and K (the total cooling step). The least sum of absolute deviations is considered as optimizing function and a robust GM(1,1) grey model is achieved. Considering the quality of results and time consumption for calculation, the optimized parameters are as follows: N=10, L=5, K=20, λ=4, α=0.9. The proposed method is applied to the study on the relationship between capacity factors of six polyaromatic hydrocarbons and the composition of the mobile phase in reverse-phase liquid chromatography and satisfactory results may be obtained even a set of abnormal data with an outlier exists.