An enhanced noise resilient K-associated graph classifier

摘要: We propose a non-parametric, noise resilient, graph-based classification algorithm.We employ relational data such as the degree of relevancy.We combine smaller components together to build larger ones.The algorithm is less sensitive than SVM and Decision Tree.The shows superior performance in presence different levels noise. In this paper, we algorithm. By modifying training phase k-associated optimal graph algorithm, proposing new labeling testing phase, introduce novel approach that robust level designing proposed method, each class dataset represented by set sub-graphs (components), extension introduced components. With enhancement, demonstrate our distinguishes between noisy non-noisy sub-graphs. Moreover, data, relevancy, with non-relational attributes, distance, for sample make Gravity formula main concept behind test various modifications tailor it arbitrary shape non-uniform scattering structure. compare method classifier, well two other well-known classifiers, namely, Tree Multi-Class Support Vector Machine. Confirmed t-Test score, on datasets from UCI repository. At 5% or higher, performs, average, 7% better 20%, multi-class SVM.