A neural network based on particle swarm optimization for detection forest fire

An Opening of forest for agriculture land which usually occurs in large areas and tend to be difficult to control. Economically for actors, make clearing land with that way is effective because it requires little cost and energy. However, the impact of clearing agricultural land by burning forests are disturbing ecological life and disturbing human life, especially in the health sector. There were a several research about forest fire has done but there are still many weaknesses and laxity of each method that is applies. Therefore, this study will create model to optimize detection and reduce the level of error in forest fire with the BPNN method and evaluated using RMSE. The dataset in this study are fine fuel moisture index (FFMC), the average rating of water content from organic matter on the surface (DMC), the average rating of water content from organic matter under the surface (DC), the figure ranking of expected/expected fire speed (ISI), Relative Humidity (RH), Wind speed (wind), Rainfall (rain) and Area, where the data consist of 517 data records. And the result, prediction value of RMSE (Root Mean Squared Error) is 37,364. Based on the analysis of testing between neural network models with neural network optimization with PSO is 34.199.