This paper introduces a new tree-based ensemble method called Extra-Trees, which randomizes both attribute and cut-point choices during tree node splitting. The method builds highly randomized trees whose structures are independent of the output values of the learning sample. The strength of randomization can be adjusted by a parameter, and the default choice of this parameter is evaluated for robustness and adaptability to specific problem conditions. The main strength of Extra-Trees lies in its computational efficiency, and a bias/variance analysis is provided to understand its performance. The paper also discusses the geometrical and kernel characterizations of the models induced by Extra-Trees. The method is evaluated on a diverse set of classification and regression problems, showing superior accuracy and computational efficiency compared to other ensemble methods like Bagging, Random Subspace, and Random Forests. The analysis of the effect of parameters and the bias/variance tradeoff further highlights the effectiveness of Extra-Trees in reducing variance and managing bias.This paper introduces a new tree-based ensemble method called Extra-Trees, which randomizes both attribute and cut-point choices during tree node splitting. The method builds highly randomized trees whose structures are independent of the output values of the learning sample. The strength of randomization can be adjusted by a parameter, and the default choice of this parameter is evaluated for robustness and adaptability to specific problem conditions. The main strength of Extra-Trees lies in its computational efficiency, and a bias/variance analysis is provided to understand its performance. The paper also discusses the geometrical and kernel characterizations of the models induced by Extra-Trees. The method is evaluated on a diverse set of classification and regression problems, showing superior accuracy and computational efficiency compared to other ensemble methods like Bagging, Random Subspace, and Random Forests. The analysis of the effect of parameters and the bias/variance tradeoff further highlights the effectiveness of Extra-Trees in reducing variance and managing bias.