Building dynamic fuzzy regression model based on convex hull algorithm and its industrial applications

Ramli, Azizul Azhar (2016) Building dynamic fuzzy regression model based on convex hull algorithm and its industrial applications. PhD thesis, Waseda University, Japan.

[img]
Preview
PDF
14Mb

Abstract

The conventional regression model was widely used in various real applications for 50 years. Precise and accurate models for prediction are important, especially for decision making purposes. Fuzzy logic is an approach to computing based on the "degrees of truth" rather than the common "true or false" Boolean logic, on which a modern computer is based. In other words, it deals with reasoning that is approximate rather than fixed or exact. Nowadays, a huge number of transactions are produced with an enormous amount of raw data. Comparing with the statistic regression method, fuzzy regression model however, requires large computation time, especially to dynamic or "on-line" data processing, because it takes all fuzzy samples as data points. The main aim of this research is to propose an innovative and efficient approach by combining the convex hull approach with the fuzzy regression technique to deal with dynamic data processing. This approach enables the computation of huge data within a realistic time frame. Three new dynamic models of the convex hull-based fuzzy regression are proposed and applied in industrial applications including granular data and switching regression problem. Specifically, well-known techniques such as conventional regression model are thoroughly extended and generalized. Statistical analysis techniques have been widely studied to improve its capability and produce effective prediction or forecasting models. However, these conventional techniques are not suitable for possibilistic or vague data processing. Then, since early 2000, some previous research studies highlighted the combination (or hybrid) approach for enhancing the computational performance and increasing the quality of produced regression models. However, most of them focus on batch data processing, which is not applicable for dynamic data analysis processes. Recently, an incipient practice called Soft Computing (SC) technology helps models and classifiers exploit tolerance for imprecision and uncertainty. The main objective of this research is to specifically create and design an algorithm for producing fuzzy regression models for dynamic data processing. A mathematical geometry facility such as Beneath-Beyond algorithm has been chosen to be combined together as one of convex hull incremental algorithm. Hence, the computational complexity and overall processing time were evaluated and successfully shown their simultaneously decrease. Therefore, decision making processes will be more efficient and well-timed.

Item Type:Thesis (PhD)
Subjects:Q Science > QA Mathematics > QA273 Probabilities. Mathematical statistics
Divisions:Faculty of Computer Science and Information Technology > Department of Software Engineering
ID Code:9166
Deposited By:Mr. Mohammad Shaifulrip Ithnin
Deposited On:03 Jul 2017 08:54
Last Modified:03 Jul 2017 08:54

Repository Staff Only: item control page