A SIMPLE LOCALLY ADAPTIVE NEAREST NEIGHBOR RULE WITH APPLICATION TO POLLUTION FORECASTING

摘要: In this paper, we propose a thorough investigation of nearest neighbor rule which call the "Symmetric Nearest Neighbor (sNN) rule". Basically, it symmetrises classical relationship from are computed points voting for some instances. Experiments on 29 datasets, most readily available, show that method significantly outperforms traditional Neighbors methods. domain interest related to tropical pollution normalization also greater potential method. We finally discuss reasons rule's efficiency, provide methods speeding-up classification time, and derive sNN reliable fast algorithm fix parameter k in k-NN rule, longstanding problem field.