Research on Improved Attribute Reduction Algorithm of Massive Incompatible Decision Data

Incompatible attributes reduction is the main method to solve the logical attribute reduction and rule of massive decision data. The reduction of incompatible equivalence relation has different alignment problem. Because the recursive algorithm is implemented in traversing the discernibility matrix, the reduction efficiency is low. According to the massive incompatible decision data, two kinds of relative attribute reduction algorithms are proposed. An incompatible decision algorithm based on equivalence class is used for relative attribute reduction of massive data. An improved information function of discernibility matrix is defined for simplifying the condition matrix in rule. Column attribute is increased for realizing the core conversion of relative attribute. Finally, the relative discernibility matrix is established to simplify the logic operation process. Experiment and simulation results show that this method can reduce the recursive computation. The easy attributes are increased, and the complex incompatible massive decision data are transformed into simple compatible decision data, and the data mining performance is improved as result.