Data Analysis for Prediction of Forest Fires

Abstract

The earth offers various resources for humanity but these resources are limited. Somethings have to be done in order to save this resources and also nature within sustainability perspective. Nature is essential for life by offering water resources, clean air and various food resources, etc. Forests have a great impact on these essential resources besides aesthetics. However, many forests were lost due to unexpected and uncontrolled forest fires especially at recent years. The main idea of this study is, an effective forest fire prediction system can help us to save forests. In this study, several basic methodologies are studied to predict the forest fires with a dataset containing meteorological data, from the literature. Some preventive actions can be taken with a successful prediction information. Also this prediction facilitates the planning of resources to cope with forest fires. For prediction study, neural networks, linear regression, decision tree and random forest methodologies are used within Python and MATLAB environment. The dataset from the literature which contains meteorological data (temperature, rain, humidity, wind) and also size of burned area data, is used in this study. Results are compared and it is realized that neural networks performs better than the other in means of accuracy value, for forest fire prediction with meteorological data.