Comparative Study based on Error Calculation in Multiple Linear Regression Coefficient for Forest Fires Prediction

Abstract

Forest fires are one of the most frequent natural disasters in Indonesia which often happens during the dry season. This causes losses in the form of threatened habitat for flora and fauna, disrupts the balance of the ecosystem, increases the potential for other natural disasters, reduce water resources, cause pollution and global warming. A land with potential high fire risk has several indications that can be used to calculate a prediction by carefully analyzing the indications. Parameters that used in this research are temperature, humidity, wind, and rain. The main proposed of this research is to know which the optimal method to find coefficient value of multiple linear regression. One commonly used prediction method is multiple linear regression method. The calculation result by using multiple linear regression method is very influenced by the determination of the coefficient value. The calculation of multiple linear regression coefficient can be done with the method of gauss-jordan, gauss-seidel and least-square. This research shows the different methods of calculating the value of multiple linear regression coefficient. The calculation of multiple linear regression coefficient in this study is known to produce the best values of Residual Standard Error (RSE) 62.726442, Mean Square Error (MSE) 3934.60648, Root Mean Square Error (RMSE) 64.6974 and Mean Absolute Percentage Error (MAPE) 310.44549 that proof it tend to be better when using the least-square method.