Research On Knowledge Acquisition Method Of Information System Based On MPI

The attributes of information systems in real life are not only diversified and high-dimensional,but also include noise,redundancy and irrelevant attributes.In order to eliminate the effects of noise on the calculation process and the final result,and solve the complexity and accuracy of data calculation,It is necessary to choose the right knowledge acquisition method.With the explosive growth of data in recent years,parallel technologies are becoming more and more important.Parallel computing can improve performance and storage space,allowing single-machine computing and storage bottlenecks to be resolved.This thesis starts from the rough set theory and applies the parallel computing idea to the knowledge acquisition method.It aims to combine the parallel computing and the rough set theory to solve the problems of the computational performance based on the complexity and accuracy of the data calculation,computational efficiency,dealing with serial algorithms in the calculation of large information systems,such as the long-playing emergence,overflow memory and downtime and other issues.The main research work is as follows:1.For the discretization of continuous data,the method of clustering continuous attributes based on MPI in parallel is proposed.First,the information system is divided into blocks,and these data blocks are allocated to each computing node.Secondly,after the data is normalized in the parallel clustering stage,the attributes are clustered based on the k-means algorithm to obtain clustering results.Finally,simulation experiments were conducted using tens of thousands to hundreds of thousands of data sets,the simulation experiment was conducted and the experimental results show that the parallel clustering discretization algorithm based on MPI greatly shortens the running time of the serial algorithm.It can be used to the process of large-scale data,and improve the computational performance and computational efficiency.2.Aiming at the problem of knowledge reduction of discretized data,this paper proposes a parallel method based on MPI to solve knowledge reduction using the resolution matrix.First,allocate the parallel tasks and calculate the equivalence classes at the same time to obtain the equivalence class results.Second,each node builds the resolution matrix in parallel according to the assigned tasks,and calculate the attributekernel.Then,the knowledge reduction results are calculated by the attribute kernel.Last,simulation experiments were conducted using hundreds to hundreds of thousands of data sets,the experimental results show that the algorithm based on MPI using the resolution matrix and the parallel knowledge reduction algorithm can keep the reduction result unchanged and solve the serial algorithm's time and space bottlenecks and handle large-scale data sets.The parallel computing method in this paper is easy to expand.If you increase the available computing resources,the calculation speed will be further improved.