A hybrid ensemble learning framework for basketball outcomes prediction

Basketball outcomes prediction is a vital technique for prospective player arrangement, injury avoidance, telecast right pricing, etc., which requires a understanding of the skill, luck, and other exterior factors of both teams. This paper presents a hybrid ensemble learning framework for basketball outcomes prediction by learning the recent status of the teams. To achieve this, we first design a new weighted combination feature for a future game by considering the latest status of the home team and the visiting team. Then, we present a hybrid ensemble framework equipped with bagging strategy and random subspace method to enlarge the diversity of the samples by learning a series of support vector machines. Finally, we develop a voting mechanism to predict the basketball outcomes. Extensive experiments have demonstrated the outperformance of our framework. (C) 2019 Elsevier B.V. All rights reserved.