Sliding Window Technique For Forest Fire Prediction
Every year, forest fire in Portugal causes large areas of land being destroyed and there are cases of death. In this research pattern discovery is being used to generate patterns of meteorological conditions in relation to area burnt of forest fire. The meteorological conditions that are being inves...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | eng eng |
| Published: |
2008
|
| Subjects: | |
| Online Access: | https://etd.uum.edu.my/559/1/Khor_Jia_Yun.pdf https://etd.uum.edu.my/559/2/Khor_Jia_Yun.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Every year, forest fire in Portugal causes large areas of land being destroyed and there are cases of death. In this research pattern discovery is being used to generate patterns of meteorological conditions in relation to area burnt of forest fire. The meteorological conditions that are being investigated are temperature, relative humidity, wind speed and rainfall. The combination of these four conditions forms the patterns that are of interest in
this research. The sliding window technique is being used to generate patterns for meteorological conditions that are significant to forest fire. The initial dataset is being transformed by changing the continuous values of the attributes into categorical values of the attributes. The patterns are then being generated through the sliding window methodology. Patterns that could not be validated are being regarded as invalid and thus are discarded while the patterns that could be validated are taken for further analysis. Patterns that are valid are then being grouped based on the burnt area associated with a pattern. The rules are then generated by transforming the categorical values into intervals and the merging of different records into the same rules. The rule generation stage produces eight distinct patterns of meteorological conditions that could predict the size of forest fire. In addition, this study showed that the sliding window technique could be used in non-temporal data. |
|---|