The most dominant source of indoor radon is the underlying soil, so the enhanced levels of radon are usually expected in mountain regions and geology units with high radium and uranium content in surface soils. Laboratory for radioactivity and dose measurement, Faculty of Sciences, University of Novi Sad has rich databases of natural radionuclides concentrations in Vojvodina soil and also of indoor radon concentrations for the region of Vojvodina, Northern Province of Serbia. In this paper we present the results of correlative and multivariate analysis of these results and soil characteristics in order to estimate the geogenic radon potential. The correlative and multivariate analysis were done using Toolkit for Multivariate Analysis software package TMVA package, within ROOT analysis framework, which uses several comparable multivariate methods for our analysis. The evaluation ranking results based on the best signal efficiency and purity, show that the Boosted Decision Trees (BDT) and Multi Layer Preceptor (MLP), based on Artificial Neural Network (ANN), are multivariate methods which give the best results in the analysis. The BDTG multivariate method shows that variables with the highest importance are radio-nuclides activity on 30 cm depth. Moreover, the multivariate regression methods give a good approximation of indoor radon activity using full set of input variables. On several locations in the city of Novi Sad the results of indoor radon concentrations, radon emanation from soil, gamma spectrometry measurements of underlying soil and geology characteristics of soil were analyzed in detail in order to verify previously obtained correlations for Vojvodina soil. •Multivariate analysis of indoor radon and soil characteristics were performed.•The multivariate methods identify the most significant variables and radon priority areas.•Radionuclides in deep layers, humus and clay content have the most impact to indoor radon.