A new navigation system troubleshooting method based on improved decision tree algorithm is proposed. Firstly, the original attributes of navigation data are sorted by information gain rate (IGR). Then the sorted attributes are clipped with neural network. Finally, the minimum decision tree is built with the clipped attributed, as well as the classification rules. Compared with the traditional decision tree algorithm, this improve decision tree picks attributes with IGR, and build the minimum decision tree directly, which can avoid the node clipping deal. The experiments indicated that this algorithm need shorter extraction time, and get higher precision.