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.

A Feature Discretization Method Based on Fuzzy Rough Sets for High-Resolution Remote Sensing Big Data Under Linear Spectral Model

“…We used FRSGA and reinforcement learning-based genetic algorithm (RLGA) [20] to optimize the discretization results Table X shows the classification accuracy of the neural network classifier after being trained by the discretization results of the above algorithms. After optimizing the discretization scheme of MFD-mvtR, the classification accuracy obtained by the confusion matrix was about 10 percentage points higher than the original, and the kappa coefficient was 0.9475.…”

“…Thus, it deals more flexibly with uncertain information introduced by mixed pixels. In our previous research work [20], we used genetic algorithm to optimize the multidimensional data discretization scheme, and determined the crossover segments and mutation points of the discretization scheme to be optimized through Qlearning mechanism, and achieved good discretization results. However, this method uses conventional genetic operations on other individuals in the population, so it is difficult to obtain high-quality individuals to cross with the discretization scheme to be optimized, which reduces search efficiency.…”

“…In remote sensing image feature discretization, neglecting the influence of the uncertainty caused by the mixed spectrum will cause errors in the data and reduce the classification accuracy after discretization [29]. With the advantage of rough set in dealing with incomplete information, the discretization method of combining rough set and genetic algorithm [20] can make full use of the powerful search ability and the known knowledge base without any prior information to obtain the minimum number of breakpoints while ensuring that the compatibility of the system is not destroyed. However, the equivalent class of rough set is an ordinary set, which is difficult to describe the fuzzy components in the data, and the accuracy obtained in the discretization of mixed spectrum is low.…”

“…In view of the above problems, we propose a discretization method called the fuzzy-rough-set-based genetic algorithm (FRSGA) for high-resolution satellite remote sensing of big data. Our innovation points to the article are as follows: (1) different from the current discretization methods, which are based on the assumption that one pixel only corresponds to the spectral information of a single object, the influence of uncertainty caused by a mixed spectrum is fully considered in our method, each category of ground object in high-resolution remote sensing image is regarded as a fuzzy set, and the Euclidean distance between pixels is calculated according to the pixel values on each band to obtain the fuzzy relation matrix on the set of pixels; (2) based on the rough set model in the original work, the membership degree of pixels to each category is determined through linear decomposition, and the individual fitness function based on a fuzzy rough model [19] is established; (3) in contrast to the conventional genetic operations in the previous work [20], chaotic mapping is employed to generate the initial population of adaptive genetic algorithm [30] to select discrete breakpoints, and MapReduce [31] is used to calculate the fitness of all individuals of the population in parallel, hence obtaining the optimal discretization scheme in the minimum time. We compared our method with mainstream discretization methods on the authentic remote sensing datasets.…”

RSFD: A rough set-based feature discretization method for meteorological data

“…Feature discretization is a key technology of intelligent data preprocessing (Chen et al, 2018). It removes redundant information by converting continuous features in meteorological data into discrete ones that are closer to the knowledge layer representation, thus reducing the system overhead and enhancing the robustness of the learning algorithm (Chen et al, 2020;Huang et al, 2020;. In addition, feature discretization can be useful for missing value imputation (Rahman and Islam, 2016).…”