In recent years, generative adversarial networks (GANs) have been proposed to generate simulated images, and some works of literature have applied GAN to the analysis of numerical data in many fields, such as the prediction of building energy consumption and the prediction and identification of liver cancer stages. However, these studies are based on sufficient data volume. In the current era of globalization, the demand for rapid decision-making is increasing, but the data available in a short period of time is scarce. As a result, machine learning may not provide precise results. Obtaining more information from a small number of samples has become an important issue. Therefore, this study aimed to modify the generative adversarial network structure for learning with small numerical datasets, starting with the Wasserstein GAN (WGAN) as the GAN architecture, and using mega-trend-diffusion (MTD) to limit the bound of virtual samples that the GAN generates. The model verification of our proposed structure was conducted with two datasets in the UC Irvine Machine Learning Repository, and the performance was evaluated using three criteria: accuracy, standard deviation, and p-value. The experiment result shows that, using this improved GAN architecture (WGAN_MTD), small sample data can also be used to generate virtual samples that are similar to real samples through GAN.

The problem of imbalanced data has a heavy impact on the performance of learning models. In the case of an imbalanced text dataset, minority class data are often classified to the majority class, resulting in a loss of minority information and low accuracy. Thus, it is a serious challenge to determine how to tackle the high imbalance ratio distribution of datasets. Here, we propose a novel classification method for learning tasks with imbalanced test data. It aims to construct a method for data preprocessing that researchers can apply to their learning tasks with imbalanced text data and save the efforts to search for more dedicated learning tools. In our proposed method, there are two core stages. In stage one, balanced datasets are generated using an asymmetric cost-sensitive support vector machine; in stage two, the balanced dataset is classified using the symmetric cost-sensitive support vector machine. In addition, the learning parameters in both stages are adjusted with a genetic algorithm to create an optimal model. A Yelp review dataset was used to validate the effectiveness of the proposed method. The experimental results showed that the proposed method led to a better performance subject to the targeted dataset, with at least 75% accuracy, and revealed that this new method significantly improved the learning approach.

The antlion optimization algorithm (ALO) is one of the most effective algorithms to solve combinatorial optimization problems, but it has some disadvantages, such as a long runtime. As a result, this problem impedes decision makers. In addition, due to the nature of the problem, the speed of convergence is a critical factor. As the size of the problem dimension grows, the convergence speed of the optimizer becomes increasingly significant. Many modified versions of the ALO have been developed in the past. Nevertheless, there are only a few research articles that discuss better boundary strategies that can increase the diversity of ants walking around an antlion to accelerate convergence. A novel exponential-weighted antlion optimization algorithm (EALO) is proposed in this paper to address slow convergence rates. The algorithm uses exponential functions and a random number in the interval 0, 1 to increase the diversity of the ant’s random walks. It has been demonstrated that by optimizing twelve classical objective functions of benchmark functions, the novel method has a higher convergence rate than the ALO. This is because it has the most powerful search capability and speed. In addition, the proposed method has also been compared to other existing methods, and it has obtained superior experimental results relative to compared methods. Therefore, the proposed EALO method deserves consideration as a possible optimization tool for solving combinatorial optimization problems, due to its highly competitive results.