Multi-agent classifier system based on heterogeneous classifier
The MAS model consists of several independents agents, and these agents has the ability to carry out a specific task and to make decisions. When working, these agents will share information with each other. Indirectly, this allows the system to get better predictions. When the constituent agents in...
| Main Author: | |
|---|---|
| Format: | Undergraduates Project Papers |
| Language: | English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/26938/ http://umpir.ump.edu.my/id/eprint/26938/ http://umpir.ump.edu.my/id/eprint/26938/1/Multi-agent%20classifier%20system%20based%20on%20heterogeneous.pdf |
| Summary: | The MAS model consists of several independents agents, and these agents has the ability to carry out a specific task and to make decisions. When working, these agents will share information with each other. Indirectly, this allows the system to get better predictions. When the constituent agents in a MAS model consist of classifiers, the resulting system is known as a multi-agent classifier system (MACS). In this project, our focus is about mutli-agent classifier system based on heterogeneous classifiers. This is because based on the previous analysis, previous MACS model that used homogeneous type of classifiers i.e., FMMs or EFMM have problem with noise effect and noise tolerance, where both classifiers have no mutant against noise. That could have a negative effect on the classification performance. In fact, learning with noise data can cause false knowledge which will be represented as noisy hyperbox in the topology of the classifier. In order to solve this problem we propose to use a heterogeneous classifiers with pruning strategy that have the ability to reduce noise effects. That could improve the MACS classification performance by overcomes the limitations of each classifier when handling different classification problems. |
|---|