A functional link neural network with modified cuckoo search for prediction tasks
The impact of temperature, relative humidity and ozone changes bring a sharp warming climate. These changes can cause extreme consequences such as floods, hurricanes, heat waves and droughts. Therefore, prediction of temperature and relative humidity is an important factor to measure the environment...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English English English |
Published: |
2017
|
Subjects: | |
Online Access: | http://eprints.uthm.edu.my/865/1/24p%20SITI%20ZULAIKHA%20ABU%20BAKAR.pdf http://eprints.uthm.edu.my/865/2/SITI%20ZULAIKHA%20ABU%20BAKAR%20COPYRIGHT%20DECLARATION.pdf http://eprints.uthm.edu.my/865/3/SITI%20ZULAIKHA%20ABU%20BAKAR%20WATERMARK.pdf http://eprints.uthm.edu.my/865/ |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | The impact of temperature, relative humidity and ozone changes bring a sharp warming climate. These changes can cause extreme consequences such as floods, hurricanes, heat waves and droughts. Therefore, prediction of temperature and relative humidity is an important factor to measure the environmental changes. Neural network, especially the Multi-Layer Perceptron (MLP) which uses Back Propagation algorithm (BP) as a supervised learning method, has been successfully applied in various problems for meteorological prediction tasks. However, this architecture has still been facing problems which the convergence rate is very low due to the multi layering topology of the network. Thus, this research proposed an implementation of Functional Link Neural Network (FLNN) which composed of a single layer of tunable weight trained with the Modified Cuckoo Search algorithm (MCS). The proposed approach was used to predict the daily temperatures, relative humidity and ozone data. Extensive simulation results have been compared with standard MLP trained with the BP, FLNN with BP and FLNN with CS. Promising results have shown that the proposed model has successfully out performed 14% percentage compared to other network models with reduced prediction error and fast convergence rate. |
---|