In information fusion system, many algorithms that possess performance characteristics are existed to solve a single problem. The usual approach in this situation is to manually select the algorithm which has the best average performance. However, this strategy has drawbacks when the whole information fusion procedure is divided into several steps. This paper presents a modeling method that uses Markov decision process to guide algorithm selection and combination with fast performance prediction and evaluative feedback. The experimental study focuses on the classic problems of target tracking in an actual information fusion system. The encouraging results reveal the potential of applying Markov decision process to algorithm management problem.