A prediction method based on improved ridge regression

Abstract

According to the problem of Multivariate Linear Regression Model is not accurate in judging linear relationship between independent variables and dependent variables in the prediction of forest fire area. The paper takes advantage of the ridge regression model to eliminate the multicollinearity, model the data and forecast the fire area. Firstly, some variables are chosen and removed, which are of unstable standardized ridge regression coefficients or stable coefficients with small absolute values. The remaining attributes are regarded as input values of a new dataset for the Support Vector Machine model. Secondly, the new dataset is divided into training set and test data, from which classification results can be obtained. Finally, the accuracy of the model is discussed based on the outcome. Experimental results indicate the method can predict the fire areas effectively.