Feature selection for high-dimensional data based on scaled cross operator threshold filtering specific memory algorithm

This paper investigates the problem of data feature selection. Based on the basic principle of wavelet threshold filtering, the threshold parameters and threshold function are selected to process the feature data. A genetic algorithm is chosen to optimize the wavelet threshold filtering algorithm, and the scaling crossover operator and threshold filtering parameters are further designed. The optimization method of this paper is compared with other algorithms in different data sets for causal feature relationship extraction comparison and classification error rate comparison. The effectiveness of the scaling crossover operator has been verified. In five benchmark synthetic datasets with a sample size of 500, the optimization method of this paper generally outperforms other algorithms in F1, Precision and Recall, and Run-time, and is able to effectively extract causal feature relationships among data. In a total of 20 comparisons of classification error rate, the optimization method in this paper won 16 times and ranked first in 4 out of 5 datasets. It is verified that the optimization method presented in this paper is effective in dealing with high-dimensional datasets. The scaled crossover operator is capable of obtaining a smaller subset of features in the dataset, demonstrating its significant role in enhancing the classification accuracy of the optimization method presented in this paper.