Bijlsma, R (2022) Forest fire spread prediction using particle swarm optimization. Master's Thesis / Essay, Artificial Intelligence.
|
Text
RuurdBijlsma_MasterThesis_ForestFirePrediction.pdf Download (6MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (85kB) |
Abstract
Wildfires can be devastating to forest ecosystems and human populations. If we could predict the spread of these fires, they could be controlled more easily. One way to achieve this is to build a simulation of wildfires that learns from real-world data. This thesis describes the theory behind such simulations, a novel simulation that aims to be accurate and fast, and a qualitative analysis of the novel simulation. We hypothesise that the novel simulation could be computationally faster than existing solutions to this problem, while retaining good accuracy. This novel simulation was successfully implemented using real-world weather, land cover, elevation, and burnt area data, with optimization being done using particle swarm optimization. The simulation achieves an area under the curve (AUC) value of 0.67 on test data when comparing simulated burnt area to real-world burnt area data, which is on par with state of the art solutions in some cases or slightly worse in others. This relatively low accuracy value could be due to a local minimum, lack of data, low quality data, or insufficient configurability of the simulation, causing poor generalisation performance. Besides reinforcement learning, the simulation could be suitable for other applications, such as helping firefighters predict the spread of wildfires.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Wiering, M.A. and Verheij, H.B. and Sabatelli, M. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 20 Apr 2022 09:31 |
| Last Modified: | 20 Apr 2022 09:31 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/26771 |
Actions (login required)
 |
View Item |