The fuzzy rough set model has been widely used to deal with the attribute reduction of real-valued data. The reduction methods of related models have also received widespread attention. Presorting attributes before performing attribute reduction is an effective method to improve reduction efficiency. However, due to the complexity of data, accurately measuring the relationships between attributes and correctly ranking them has become a challenge. This paper defines a new overlap degree (NOD) by analyzing the overlap degree and distance calculation of common data types, and then uses the new overlap degree to pre sort attributes and propose corresponding attribute reduction methods. Finally, the effectiveness of the algorithm is verified through examples and experiments.