A New Method for Spectral Wavelength Selection Based on Multiple Linear Regression Combined with Ant Colony Optimization and Genetic Algorithm

Wavelength selection is one of the key steps in quantitative spectral analysis, which reduces computation time while also improving prediction accuracy model. In this paper, we propose a wavelength algorithm based on ant colony optimization (ACO), absolute value regression coefficient multiple linear (MLR) model used as basis for evaluating importance wavelengths, and after full MLR modeling initial pheromone (MLR-ACO). each iteration, corresponding to individual with highest fitness update. The crossover operator introduced MLR-ACO (MLR-ACO-GA), individuals top 100 values are population genetic (GA). A selected frequency wavelengths greater than threshold among calculated. number coarse interval points generated according frequency, operation performed at points. Fine randomly within interval, fine operations exploit potential combining excellent other much possible. can well solve problem traditional ACO scarcity, MLR-ACO-GA avoid falling into local optimum certain extent be more flexible give play advantages MLR-ACO.