Comparing support vector regression and random forests for predicting malaria incidence in Mozambique
Orlando P. Zacarias, Henrik Boström · 2013
Accurate prediction of malaria incidence is essential for the management of several activities in the ministry of health in Mozambique. This study investigates the comparison of support vector machines (SVMs) and random forests (RFs) for this purpose. A dataset with records of malaria cases covering the period 1999-2008 was used to evaluate predictive models on the last year when developed from one up to nine years of historical data. Mean squared error (MSE) was used as the performance metric. The scheme for estimating variable importance commonly employed for RFs was also adopted for SVMs. SVMs developed from two years of historical data obtained the best prediction accuracy. Hence, if we are interested in predicting the actual number of malaria cases the support vector machines model should be chosen. In the analysis of variable importance, Indoor Residual Spray (IRS), the districts of Manhiça and Matola and month of January turned out to be the most important predictors in both the SVM and RF models.