基于区分矩阵的启发式属性约简算法

由于大量等价类元素的存在，同一等价类中的记录与其他非该等价类中的记录相比较将会产生大量空元素及重复元素，使得构造区分矩阵需要耗费大量的时间与空间。因此以信息向量为工具处理等价类，改进了区分矩阵的构造过程，有效地提高了构造区分矩阵的时空间效率；其次，利用属性频度为启发信息，给出了一种基于区分矩阵的启发式属性约简算法；最后，利用恒星天体光谱数据集，实验验证了算法的有效性。

Heuristic algorithm of attribute reduction based on discernibility matrix

Comparing the records of one equivalence class with the records which belong to other equivalence classes will product lots of empty elementS and repeating elements, so that construction of the discernibility matrix will spend too much time and space. In this paper，firstly,the discernibility matrix was constructed by using information vector as a tool to deal with equivalence classes, thus the efficiency of space complexity and time complexity of constructing discernibility matrix was improved. Secondly, a heuristic algorithm of attribute reduction based on the discernibility matrix was presented by using attribute frequency as heuristic information. Finally, the experimented results validate the algorithm with celestial body spectra.