Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


45641.pdf (2.43 MB)

ETD Abstract Container

Abstract Header

Classification of objects using a Cascading Genetic Fuzzy System

Kunhambu Nair, Dipin
http://orcid.org/0000-0002-8960-1616
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1692273461483578

Abstract Details

2023, MS, University of Cincinnati, Engineering and Applied Science: Mechanical Engineering.
The accurate classification of data holds immense value across numerous domains. Although AI techniques used for classification are very accurate, they are often seen as "black boxes" that lack transparency and interoperability to explain the model. To address this limitation, this research explores the application of a cascading genetic-fuzzy system to enhance the explainability of the classification process. The focus of this master's thesis is the classification of two distinct types of rice and raisin varieties. Leveraging computer vision techniques, grayscale images of the rice samples are converted and processed to extract relevant attributes. These attributes serve as inputs to a fuzzy inference system (FIS) employing a 7-input 2-output architecture. To optimize the FIS's performance, a Genetic Algorithm (GA) is employed to fine-tune the membership functions for each input and output, as well as the rule base. The proposed cascading genetic-fuzzy system is rigorously evaluated and compared against existing methods to ascertain its effectiveness. By incorporating genetic algorithms within fuzzy systems, the approach strikes a balance between accuracy and transparency, allowing users to gain deeper insights into the classification process. Notably, the implemented system achieves an impressive accuracy of 95% and 87% on a validation set of rice and raisins respectively compared to the popular AI techniques such as Linear Regression, Standard Vector Machine, and Multi-Layer Perceptron. It is to note that cascading genetic-fuzzy system outperforms other models with more explainability. Through this master's thesis, we aim to advance the development of more interpretable and accurate classification models for real-world scenarios. By highlighting the importance of incorporating explainability into AI techniques, this research contributes to the overall understanding and utilization of transparent AI systems. The findings underscore the necessity of model explainability in fostering trust and confidence in decision-making processes driven by AI models.
Manish Kumar, Ph.D. (Committee Chair)
Kelly Cohen, Ph.D. (Committee Member)
Sam Anand, Ph.D. (Committee Member)
70 p.
Mechanical Engineering
Fuzzy; Genetic Algorithm; Explainable AI; Classification; Rice; Raisin

Recommended Citations

Citations

  • Kunhambu Nair, D. (2023). Classification of objects using a Cascading Genetic Fuzzy System [Master's thesis, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1692273461483578

    APA Style (7th edition)

  • Kunhambu Nair, Dipin. Classification of objects using a Cascading Genetic Fuzzy System. 2023. University of Cincinnati, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1692273461483578.

    MLA Style (8th edition)

  • Kunhambu Nair, Dipin. "Classification of objects using a Cascading Genetic Fuzzy System." Master's thesis, University of Cincinnati, 2023. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1692273461483578

    Chicago Manual of Style (17th edition)

ucin1692273461483578
45
© 2023, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.
Release 3.2.12