Graph Similarity Metric for Modifying K Nearest Neighbor for Classifying Texts

This article proposes the modified KNN (K Nearest Neighbor)algorithm which receives a graph as its input data and is applied tothe text categorization. The graph is more graphical forrepresenting a word and the synergy effect between the textcategorization and the word categorization is expected by combiningthem with each other. In this research, we propose the similaritymetric between two graphs representing words, modify the KNNalgorithm by replacing the exiting similarity metric by the proposedone, and apply it to the text categorization. The proposed KNN isempirically validated as the better approach in categorizing textsin news articles and opinions. In this article, a word is encodedinto a weighted and undirected graph and it is represented into alist of edges.