A Study of a Support Vector Machine Algorithm with an Orthogonal Legendre Kernel According to Neutrosophic logic and Inverse Lagrangian Interpolation

Abstract

The decision-making process is greatly affected by the data collection stage. If the data collection process is not well controlled, i.e. there is some data lost due to the poor quality of the devices used or the lack of accuracy in the data entry process...etc., this will affect the work of the SVM algorithm, which is considered one of the best. Most of the workbooks suffer from the problems of missing and anomalous data. In this paper, we propose a method to treat the missing and anomalous data by reshaping the data set defined by the classical method into the neutrosophical data set by calculating the amount of true T, false F, and neutrality I in the neutrosophical set using inverse Lagrangian interpolation. We noticed the superiority of our proposed method for processing missing data over the method of [21], then we trained a support vector machine algorithm with orthogonal legender kernel on a breast cancer dataset taken from the Statistics Department of Al-Bayrouni Hospital in Damascus, where the proposed algorithm achieved a classification accuracy of 97%. The reason we chose a support vector machine classifier with an orthogonal legender kernel has two goals: the first is to eliminate the repetition of support vectors in the feature space. The second is to solve the problem of non-linear data distribution.