Research on Filtering System of Harmful Information on Network Based on K Nearest Neighbor Algorithm

It is the difficult issue how to classify information accurately in the network harmful information filtering, while the K Nearest Neighbor(KNN) classification method have been shown to perform well for pattern classification in many domains. This paper presents a method of network harmful information filtering based on KNN, and improves the classification efficiency by eliminating training samples that may cause misclassification. The experiment shows that the improved system´s precision and the recall-precision have been enhanced, and classification time-consuming also has the obvious reduction.