Please use this identifier to cite or link to this item: http://hdl.handle.net/10603/462837
Title: Experimental Study on Machine Learning and Metaheuristic Techniques in Dimensionality Reduction for Data Classification
Researcher: Das, Himansu
Guide(s): Behera, H S and Naik, Bighnaraj
Keywords: Computer Science
Computer Science Software Engineering
Engineering and Technology
University: Veer Surendra Sai University of Technology
Completed Date: 2021
Abstract: Development of new technology leads to tremendous growth of data. It is a great challenge to analyze and process such a vast quantity of data. The exploration of meaningful information from such huge data is really a challenging task, without any automated system. These data contain a large amount of features that may gain computational cost of the problem. The increasing dimensions of the features could also negatively impact on the classification process due to its superfluous, extraneous, and noisy features, increases the dimension of the problem. It ultimately leads to an increase in the computational time and decreases performance of the model. In such situations, dimensionality reduction plays a crucial role to reduce the dimension of such features. Dimensionality reduction removes the irrelevant, redundant, and noisy features without loosing much intended information. It simply reduces the number of features by considering only relevant features using feature reduction and feature selection techniques. Feature reduction techniques are used to transform the input feature into its equivalent new features by filtering out the most irrelevant, redundant, and noisy information. However, feature selection uses a search technique to select a subset of original features that can effectively describe the data. The selection of most relevant features from the high dimensional data is an NP-hard problem. It may increase the possible solution space exponentially with the increase in number of features. Therefore, it is essential to reduce the dimension of the features. It will ultimately increase the performance and decrease the computational cost of the problem. The objective of the thesis is two-fold: (i) firstly, design of new hybrid classification models by considering several feature reduction techniques such as principal component analysis and linear discriminant analysis along with various neuro-fuzzy models such as class belongingness neuro-fuzzy and linguistic neuro-fuzzy.
Pagination: 
URI: http://hdl.handle.net/10603/462837
Appears in Departments:Department of Computer Science and Engineering and IT

Files in This Item:
File Description SizeFormat 
01_title.pdfAttached File183.43 kBAdobe PDFView/Open
02_prelim pages.pdf356.27 kBAdobe PDFView/Open
03_content.pdf122.09 kBAdobe PDFView/Open
04_abstract.pdf114.83 kBAdobe PDFView/Open
05_chapter 1.pdf204.01 kBAdobe PDFView/Open
06_chapter 2.pdf2.1 MBAdobe PDFView/Open
07_chapter 3.pdf1.52 MBAdobe PDFView/Open
08_chapter 4.pdf2.4 MBAdobe PDFView/Open
09_chapter 5.pdf1.78 MBAdobe PDFView/Open
10_chapter 6.pdf2.02 MBAdobe PDFView/Open
11_chapter 7.pdf2.14 MBAdobe PDFView/Open
12_chapter 8.pdf447.03 kBAdobe PDFView/Open
13_annexures.pdf155.55 kBAdobe PDFView/Open
80_recommendation.pdf605.12 kBAdobe PDFView/Open
Show full item record


Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).

Altmetric Badge: