Research On Incremental Knowledge Reduction Algorithm Based On Binary Discernibility Matrix

Knowledge reduction is one of the core contents,which can ensure the classification and decision-making capacity,to eliminate the redundant information in database.With the data increasing in an unpredictable speed in real life,in the case of insertion new data,it is time consuming and resource wasting to recalculate the whole information system to remove redundant information.Therefore,it is significant and meaningful to develop incremental attribute reduction algorithm which can update the result of attribute reduction by combining the new data with the last calculation results.For the problems that the binary discernibility matrix has large storing space consuming and how to be effectively used to calculate the knowledge reduction incrementally,we explore the incremental knowledge reduction algorithm based on binary discernibility matrix.The major study context in our paper are described as follow.(1)We explore the incremental attribute reduction algorithm based on the binary discernibility matrix in complete information system.To reduce the storage space of binary discernibility matrix,we proposed a method to compress binary matrix,simplifying the binary discernibility matrix storage space from |C|+1 column to 3.Through dynamically updating the binary discernibility matrix to incrementally get core.According to the core,we proposed an incremental attribute reduction algorithm based on binary discernibility matrix.(2)We explore the incremental attribute reduction algorithm for group dynamic data based on binary discernibility matrix is developed in complete information system.According to that the dynamic data is a single object or group objects,different branches can be chosen to update the binary discernibility matrix.The attribute core set can be easily obtained based on the updated binary discernibility matrix,on the basis of which,an incremental reduction algorithm for group dynamic objects is designed.(3)We explore the incremental attribute reduction algorithm based on the binary discernibility matrix in incomplete information system.For getting the attribute reduction incrementally,the tolerance class needs to be computed firstly.For the purpose of speeding up the tolerance class calculation,an improved static algorithm with rapidity and stability is developed firstly,followed by a novel incremental algorithm,which can update rapidly the tolerance class when a new object is coming.On the basis of the obtained tolerance class,an incremental attribute reduction algorithm based on binary matrix in incomplete information system by updating the binary matrix is proposed.(4)We explore how to use the incremental attribute reduction algorithm to extract the feature of the photovoltaic power generation prediction data.The decision table of PV power forecasting model is established based on the collected data and then the data in the decision table is discretized.When the new data is coming,the incremental attribute reduction algorithms in this paper are used to extract the feature attributes of the data set.Based on the feature attributes,the neural network algorithm is used to train the data of feature attributes and predict future photovoltaic power generation.