Feature and instances selection for nearest neighbor classification via cooperative PSO

Data reduction is an essential task in the data preparation phase of knowledge discovery and data mining (KDD). The reduction method contains two techniques, namely features reduction and data reduction which are commonly applied to a classification problem. The solution of data reduction can be viewed as a search problem. Therefore, it can be solved by using population-based techniques such as Genetic Algorithm and Particle Swarm Optimization. This paper proposes the integration of feature reduction and data reduction for Nearest Neighbor (NN) classification using Cooperative Binary Particle Swarm Optimization (CBPSO). This method can overcome the limitation of using the Nearest Neighbor (NN) classifier when dealing with high dimensional and large data. The proposed method is applied to 14 real world dataset from the machine learning repository. The algorithm´s performance is illustrated by the corresponding table of the classification rate. The experimental results demonstrate the effectiveness of our proposed method.