Reduction and Dynamic Discretization of Multi-attribute Based on Rough Set

In majority of approaches of multi-attributes discretization, the results with a large number of break points always tend to make irrational and redundant. To this issue, this paper presents a dynamic multi-attribute algorithm based on rough set. This algorithm performs reduction to the attributes from the decision-making table through signification which generated by conditional entropy, then it takes the grey correlation conception to order the attributes ascendingly after the reduction. The multi-attributes are dynamically discretized with the idea of frequency surveyed breakpoint according to the second order and quantized so as to gain the decision-making table. The results show that the method not only reduces the redundancy of breakpoints, but also improves its rationality and discrete accuracy comparing with related studies.