Open AccessOutlier detection in neutrosophic sets by using rough entropy based weighted density method
Author(s) -
Sangeetha Tamilarasu,
Amalanathan Geetha Mary
Publication year - 2020
Publication title -
caai transactions on intelligence technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.613
H-Index - 15
ISSN - 2468-2322
DOI - 10.1049/trit.2019.0093
Subject(s) - scope (computer science) , outlier , indeterminacy (philosophy) , computer science , data mining , entropy (arrow of time) , information retrieval , citation , data science , artificial intelligence , theoretical computer science , mathematics , world wide web , epistemology , physics , quantum mechanics , philosophy , programming language
Neutrosophy is the study of neutralities, which is an extension of discussing the truth of opinions. Neutrosophic logic can be applied to any field, to provide the solution for indeterminacy problem. Many of the real‐world data have a problem of inconsistency, indeterminacy and incompleteness. Fuzzy sets provide a solution for uncertainties, and intuitionistic fuzzy sets handle incomplete information, but both concepts failed to handle indeterminate information. To handle this complicated situation, researchers require a powerful mathematical tool, naming, neutrosophic sets, which is a generalised concept of fuzzy and intuitionistic fuzzy sets. Neutrosophic sets provide a solution for both incomplete and indeterminate information. It has mainly three degrees of membership such as truth, indeterminacy and falsity. Boolean values are obtained from the three degrees of membership by cut relation method. Data items which contrast from other objects by their qualities are outliers. The weighted density outlier detection method based on rough entropy calculates weights of each object and attribute. From the obtained weighted values, the threshold value is fixed to determine outliers. Experimental analysis of the proposed method has been carried out with neutrosophic movie dataset to detect outliers and also compared with existing methods to prove its performance.