Research On Three-way Decisions Incremental Learning Theory And Its Application Based On Forgetting Mechanism

The data in the dynamic environment usually presents the characteristics of continuous growth and dynamic update.The new data generated in real time will lead to the dynamic change of the original knowledge structure.The three-way decisions incremental learning model makes full use of the historical training results in learning,and reduces the time and space requirements of the knowledge updating method when the data grows rapidly through the local learning of the new part of the data.With the accumulation of more old knowledge in the incremental model,the classification performance of the system is disturbed,resulting in the problem of concept drift in the process of increasing system data.In recent years,incremental learning model has become a key research topic in dynamic three-way decisions,and has made significant achievements in both theoretical and practical fields,but there are few related researches on concept drift in incremental model.Based on the theory of forgetting mechanism,this paper designs an algorithm model with data active forgetting as the core,and proposes a three-way decisions active forgetting algorithm.Based on the calculation method of active forgetting,two kinds of three-way decisions incremental learning optimization calculation methods are proposed.The specific research work includes:(1)Based on the three-way decisions model of probabilistic rough set,by analyzing the data change mode under the dynamic three-way decisions environment,the change of the original decision rule condition probability of information system in the data active forgetting calculation is deduced,and based on this,the forgetting renewal principle of the three-way decisions approximation of probabilistic rough set is given,and the fast forgetting calculation of the data to be forgotten is realized.Finally,a comparative experiment is carried out on the public data set.The experimental results show that compared with the classical algorithm,the active forgetting algorithm has a huge advantage in running time cost.(2)In order to effectively deal with the concept drift in the three decision incremental learning model,based on the active forgetting algorithm,the backtracking idea and heuristic optimization algorithm are introduced into the optimization calculation of the three decision incremental learning,and two kinds of three-way decisions incremental learning optimization calculation algorithms based on the active forgetting calculation are proposed to explore the best performance of the three-way decisions incremental learning model.Finally,a comparative experiment is carried out on the artificial data set.The experimental results show that the two algorithms can effectively deal with the concept drift problem in the incremental model.(3)In order to solve the problem of conceptual drift in oil temperature analysis in the production of State Grid.Based on the example of oil temperature prediction and analysis,a series of algorithm flow is designed,which shows the general application process of two kinds of three-way decisions incremental learning algorithms with forgetting mechanism.The results of several comparative experiments show that the three-way decisions incremental learning algorithms with forgetting mechanism have great advantages in oil temperature analysis.