A New Feature Analysis Approach to Selecting Channels of EEG for Fatigue Driving

Algorithm 1

ReliefF.

Input: Training set ; number of samples ; number of nearest neighbor samples ;

Output: Feature weights for each feature;

1: Set all feature weights to 0 and set the empty set ;

2: For to :

    Select a sample from , randomly.

    Find the nearest neighbors of from the similar sample set of , and the nearest neighbors from each of the sets of samples of different classes;

3: For to (all features):

   This is calculated as shown in equation 10.

   Where denotes the difference between samples and on feature , and denotes the nearest neighbor sample in class .

   This is calculated as shown in equation 11;

4: Return feature weights for each feature.