An improved decision tree algorithm based on hierarchical neighborhood dependence

Neighborhood rough sets (NRS) is widely used in various fields with good adaptability, and its related information measurement plays an important role in uncertainty analysis. In the existing research, although the three-layer granular structure of neighborhood decision system (NDS) has been proposed, there are not many related research on it, especially in the uncertainty measurement. Therefore, this paper firstly deeply discusses the expression of neighborhood dependence in the three-layer granularity structure of neighborhood decision system, as well as its relationship in three levels and some related properties. Secondly, considering the influence of the fixed neighborhood radius(Nr) on the neighborhood model, we define the adaptive neighborhood radius by using the standard deviation and the neighborhood mean. Finally, we use the neighborhood dependence at the macro top level as the split node measurement function to construct the decision tree. Our experimental results show the reliability of the algorithm.