Dimensionality Reduction based on Uncertain Graph Model

Classification in machine learning is done by many factors which called attributes. The higher the number of features, the more difficult it becomes to visualize the training set and then work on it. Sometimes, most of these features are related to each other and are therefore considered redundant features. This is where Dimensionality Reduction (DR) algorithms come into play. In machine learning and statistics, dimensionality reduction is the process of reducing the number of supervised random variables by obtaining a set of main variables. Dimensionality reduction can be divided into feature selection and feature extraction. This paper proposes a new Dimensionality reduction algorithm in the feature selection category using Pearson correlation of attributes and making uncertain graph models. The proposed model can be done for any number of features with increasing the classification performance compared with filter and wrapper strategies.