An Optimal Feature Selection Process using Roughset Theory in High Dimensional Data Classification

Computer Science - Data Mining

This research entitled “AN OPTIMAL FEATURE SELECTION PROCESS USING ROUGHSET THEORY IN HIGH DIMENSIONAL DATA CLASSIFICATION” incorporates information theory, which is the process of deriving the information from the feature selection from the unsupervised dataset. Feature Selection is the application of data mining techniques to discover patterns from the micro array datasets. Finding the best features that are similar to a test data is challenging task in current trend. To discover the significance features have more frequent change in the structural information, which involves feature dimensionality reduction, linked to one another and elimination of non-structural information. This research presents a framework for discovering best feature selection from unsupervised datasets. By aligning the relevant features from the datasets and by using the matching sequence or its frequency of match, the searching between the data features are determined. The proposed research work presents a new approach to measure the features (attributes) in micro array datasets using the methodologies namely, data cleaning, Adaptive Relevance Roughset Feature Discovery, minimal-Redundancy-Maximal-Relevance (mRMR) and classification. Data feature selection and dimensionality reduction is characterized by a regularity analysis where the feature values correspond to the number times that term appears in the dataset. The relevance Roughset feature discovery method gives a useful measure is used to find the similarity features between data points are likely to be in terms of their features property. Despite the usefulness of searching measures in these applications, accurately measuring the similarity between the features or attributes remains a challenging task. Some of the challenges faced in finding the best feature selection include positive, negative and inconsistency. This research proposes an enhanced relevance rough set based classification method to estimate the feature searching is measured using minimal redundancy optimization method corresponding micro array data. Each feature contains objective function and their own description which is used to identify the type of datasets. Initially, the total numbers of features are identified to enhanced feature selection of the datasets where the terms of match between the features are identified with help of classification algorithms.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.