Machine Learning Regression Techniques to Predict Burned Area of Forest Fires

摘要

The study presents the implementation of machine learning regression techniques to predict burned areas of forest fires. The data set used in this paper is presented in UCI machine learning repository that consists of climatic conditions and physical factors of the Montesinhopark in Portugal. Linear regression, ridge regression and lasso regression algorithms are used in the process of prediction. Accuracy score, Mean Absolute Error (MAE), Median Absolute Error (MDAE) and Mean Squared Error (MSE) were calculated. The size of the data set is 517 entries and the number of features for each row is 13. In this study the three algorithms are applied using two versions. In the first version, all features of the data set were included and in the second version, 70% of the features were included. In both versions, the training set is 70% of the data set and the test set is 30% of the data set. The accuracy of linear regression algorithm is 100%, thus it gives more accuracy than ridge regression and lasso regression algorithms in both versions.