As is rapidly becoming clear, data science increasingly permeates many aspects of life. Educational research recognizes the importance and complexity of learning data science. In line with this imperative, there is a growing need to investigate the factors that influence student performance in data science tasks. In this paper, we aimed to apply machine learning algorithms to predict students' data science performance using a set of features related to problem-solving behavior logged by the "DaTu" system. In terms of predictive power, our results indicate that AdaBoost achieved the best performance. The use of AdaBoost in this context highlights the potential of ensemble methods to enhance predictive accuracy. The key contribution of this study lies in its demonstration of how machine learning can uncover the underlying relationships between students' problem-solving behaviors and their proficiency in data science tasks. This knowledge has implications for informing the design of adaptive learning environments that cater to individual learners' needs.