Research on Feature Selection in Near-Infrared Spectroscopy Classification Based on Improved Adaptive Genetic Algorithm Combined with ReliefF

In order to solve the problem of feature selection and the low classification accuracy in high-dimensional data, a feature selection algorithm ReliefF-IAGA which is based on ReliefF and IAGA(improved adaptive genetic algorithm) is proposed in this paper. The algorithm obtains the feature importance score using ReliefF feature selection algorithm and eliminates the irrelevant features. Then, the optimal feature subset is obtained by using IAGA. In order to verify the effectiveness of the algorithm, the tobaccos origin classification model was constructed by using the tobaccos from different places of origin. Simulation results show that compared with other methods, the optimal feature subset obtained by the proposed method is more relevant, improving the classification accuracy of the origin of tobaccos.