Reinforcement Learning-Based Genetic Algorithm in Optimizing Multidimensional Data Discretization Scheme

Abstract

Feature discretization can reduce the complexity of data and improve the efficiency of data mining and machine learning. However, in the process of multidimensional data discretization, limited by the complex correlation among features and the performance bottleneck of traditional discretization criteria, the schemes obtained by most algorithms are not optimal in specific application scenarios and can even fail to meet the accuracy requirements of the system. Although some swarm intelligence algorithms can achieve better results, it is difficult to formulate appropriate strategies without prior knowledge, which will make the search in multidimensional space inefficient, consume many computing resources, and easily fall into local optima. To solve these problems, this paper proposes a genetic algorithm based on reinforcement learning to optimize the discretization scheme of multidimensional data. We use rough sets to construct the individual fitness function, and we design the control function to dynamically adjust population diversity. In addition, we introduce a reinforcement learning mechanism to crossover and mutation to determine the crossover fragments and mutation points of the discretization scheme to be optimized. We conduct simulation experiments on Landsat 8 and Gaofen-2 images, and we compare our method to the traditional genetic algorithm and state-of-the-art discretization methods. Experimental results show that the proposed optimization method can further reduce the number of intervals and simplify the multidimensional dataset without decreasing the data consistency and classification accuracy of discretization.