Quantitative precipitation analysis and offline gui using neural network system
This project discovers the implementation of Artificial Neural Network (ANN) for forecasting weather based on past relevant data. Neural network is constructed using empirical network architecture and (17) training types. They are such as BFGS quasi-Newton backpropagation, Cyclical order incremental...
Saved in:
| Main Author: | |
|---|---|
| Format: | Undergraduates Project Papers |
| Language: | English |
| Published: |
2009
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/1958/1/Siti_Nursyuhada_Mahsahirun_%28_CD_5362_%29.pdf http://umpir.ump.edu.my/id/eprint/1958/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Malaysia Pahang |
| Language: | English |
| Summary: | This project discovers the implementation of Artificial Neural Network (ANN) for forecasting weather based on past relevant data. Neural network is constructed using empirical network architecture and (17) training types. They are such as BFGS quasi-Newton backpropagation, Cyclical order incremental training w/learning functions, Levenberg-Marquardt backpropagation, Resilient backpropagation and others. The ANN has been trained using 2008 weather data and tested with data year 2009. As result, the system has successfully generating accuracy up to 78.69% for quantitative precipitation (QP) prediction. Analysis on time consumption of all those training types is made and shows that Resilient backpropagation with 1.92s training time consumption is the fastest and Cyclical order incremental training w/learning functions with 463.215s is the slowest. This project concluded that ANN is an alternative method in controlling and understanding the way of non-linear set of data and variables to become mutually correlated with each other. It is a powerful yet significant method in embedding intelligent system into application for meteorological tools. |
|---|