Researched For Continuous Valued Attribute Reduction Algorithm Based On Rough Theory

As mankind has entered the information age, a large amount of data information was produced in daily life. The information is all in a mess, the information has a lot of redundancies, so social activities need a tool that can distinguish the valueless data and get the useful decision rules for artificial decisions. For the Rough Set has many advantages in dealing with this problems, it has caused wide public concern and been widely used.Classic rough set model is built in strict equivalence relations which are on the basis of equivalence partitioning, so the processing object is limited to discrete data types. However, large amounts of data generated from the production work in real life is a continuous value, it needs further explored. Firstly, this paper presents the greedy algorithm and the discretization algorithm based on information entropy. It analyzes the advantages and disadvantages of these two algorithms. On this basis, we particularly analyzed the discretization algorithm of continuous data which based on attribute importance, and provided a continuous data decision table to verify the performance of these algorithms. Then we give the calculation method that processes the data in the decision table immediately. During this process we picked up the decision-making rules and the reduction of relative attribute. Combined with fuzzy rough set, we set an allowable error for system to calculate the maximal tolerance class and the discernibility function. In the end, we get the decision rules and the attribute reduction. Finally, we used a concrete decision table to verify the effectiveness of the algorithm.At the same time, we have dealt with the incomplete data table which exists missing data and data omission. We proposed expansion distinguish matrix algorithm for continuous--type decision-making table. Using this algorithm, we gained the relative reduction and nuclear properties through the use of logical operation between the elements of the matrix, and verified the effectiveness of the algorithm by specific decision-making table. Finally, we changed the incomplete data table into a complete data table by data filling algorithm, then proposed the reduction algorithm based on tree structure. The process of this algorithm is easy to understand, and can get the reductions of all the attributes in the decision tables.