Feature selection using population random search with memory

In this paper, we introduce a new method for solutions to feature selection problems. We develop a technique based on population random search with memory. The concept underlying the proposed method is a combination of random and heuristic search strategies. The solution is represented as a binary vector, the dimension of which is determined by the number of features in the data set. The generation of new solutions is carried out randomly using normal and uniform distribution. The heuristic is formulated as follows: the chance of a sign to get into the next generation is proportional to the frequency of the presence of this feature in previous best solution. The proposed method has been tested on several data sets from the KEEL repository. Comparison with other state-of-the-art feature selection methods demonstrates the competitiveness of the proposed.