Error-Aware Markov Blanket Learning for Causal Feature Selection

Usage

"EAMB.m" is main function.

Note that the current code has only been debugged on Matlab (2018a) with a 64-bit Windows system.

---

function [MB,time] = EAMB(Data,data_type,target,alpha,k_or)

INPUT:

Data is the data matrix, and rows represent the number of samples and columns represent the number of nodes. If Data is a discrete dataset, the value in Data should start from 1.

data_type='dis' represents discrete data, and data_type='con' denotes continues data.

target is the index of target node.

alpha is the significance level, e.g., 0.01 or 0.05.

k_or is recall coefficient, and k_or is greater than or equal to 0 and less than or equal to 1. Generally, k_or is taken as [0.05, 0.25]. The higher the dimension of the dataset, the smaller the value of k_or.

OUTPUT:

MB is the Markov blanket of the target.

time is the runtime of the algorithm.

Example

clear;
clc;
load('./data/leukemia/data_labels.mat');
load('./data/leukemia/cv10_indices.mat');

% Cross-validation
for i = 1:10
    test_indices = (indices == i); train_indices = ~test_indices;

    Data = horzcat(data(train_indices,:),labels(train_indices));
    Data = Data+1;
    [~,p] = size(Data);

    [features,~] = EAMB(Data, 'dis', p, 0.01, 0.05); % The p-th variable is the label variable.
end

References

[1] Guo X, Yu K, Cao F, et al. Error-aware Markov blanket learning for causal feature selection[J]. Information Sciences, 2022, 589: 849-877.