Abstract:

The forest fire danger composite index(FI) is constructed by normalized difference vegetation index (NDVI) and the relative humidity. Support vector machine (SVM) methods used for forest fire forecasting prediction test. Using meteorological data and secondary planet remote sensing data, the classification illation model and regression illation model of forest fire were based. The forecasting test was progressed too. The results show that classification illation model has favorable prediction ability, its prediction effect is superior to successive regression technique. But there is little difference in prediction effect between the regression illation model and successive regression technique.

Key words: forest fire integrated index(FI), support vector machine(SVM), prediction of forest fire disaster, classification model, regression estimation