Algorithmic Study of Gibbs Sampling Selection Based on Information Criterion

Feature selection is a key research direction in the current big data era, which can effectively reduce the dimension of data, simplify the time of model training and improve the prediction effect, attracting the attention of researchers in the field of statistics and informatics. In recent years, with the expansion of the application scope and field, the research on the feature selection method gradually approaches to the comprehensive index of time and efficiency. This paper applies the idea of Gibbs sampling based on the random search strategy, and improves the feature selection method based on the Gibbs sampling algorithm, analyzes the frequency of each feature as the important degree value of the corresponding feature, and selects a suitable threshold value to realize the feature selection according to the size of the important degree value.