TY - EJOU AU - Li, Juan TI - An Improved K-nearest Neighbor Algorithm Using Tree Structure and Pruning Technology T2 - Intelligent Automation \& Soft Computing PY - 2019 VL - 25 IS - 1 SN - 2326-005X AB - K-Nearest Neighbor algorithm (KNN) is a simple and mature classification method. However there are susceptible factors influencing the classification performance, such as k value determination, the overlarge search space, unbalanced and multi-class patterns, etc. To deal with the above problems, a new classification algorithm that absorbs tree structure, tree pruning and adaptive k value method was proposed. The proposed algorithm can overcome the shortcoming of KNN, improve the performance of multi-class and unbalanced classification, reduce the scale of dataset maintaining the comparable classification accuracy. The simulations are conducted and the proposed algorithm is compared with several existing algorithms. The results indicate that the proposed algorithm can obtain higher classification efficiency and smaller search reference set on UCI datasets. Furthermore, the proposed algorithm can overcome the shortcoming of KNN and improve the performance of multi-class and unbalanced classification. This illustrates that the proposed algorithm is an expedient method in design nearest neighbour classifiers. KW - KNN; TPKNN; unbalanced class patterns; tree structure; tree pruning; penalty parameter DO - 10.31209/2018.100000003