Correlation-based K-Nearest Neighbor Algorithm

摘要

Data classification is an important task of data mining, and developing high-powered classification algorithm is one of the key procedures for data mining. This paper provides an improved K-nearest neighbor algorithm—correlation-based K-nearest neighbor algorithm. This new algorithm makes data classification base on the correlation calculation, and uses a modified probability to improve the computational speed and prediction accuracy. The experimental results show that the correlation-based K-nearest neighbor algorithm is more suitable to classify massive high-dimensional data sets while comparing with traditional K-nearest neighbor algorithm.