Research on Hybrid Feature Selection Method Based on Iterative Approximation Markov Blanket
Algorithm 1
CI_AMB algorithm.
Input: dataset , samples, features
K//the number of feature subsets divided
Output: optimal feature subset , samples, features
Begin
Phase 1: filtering irrelevant features
For I = 1 to m: //MIC calculation
Standardize and ;
Calculating the MIC score value for each feature in ;
End
According to the evaluation index RMSE, the filtered candidate feature subset is determined, and the candidate feature subset of the dimension is arranged in ascending order;
Then, the selected candidate feature subset sequences are divided: ; //Divided into K shares
; //Initialize the optimal feature subset to be empty
Phase 2: eliminating redundant features
Performing redundancy analysis on the first feature subset using the AMB method and filtering out nonredundant features to join ;
For I = 2 to K: //Iterative AMB
; //Add the current optimal feature subset to the next partition subset
; //Update the list of optimal features using the AMB method, and finally
End
Construct a regression model and verify and evaluate the validity and reliability of the model;
