Deep Learning-based Test Data Augmentation Technology
Xiangwei Kong, Haidong Wu, Huiyang Hu · 2023
In recent years, deep learning has developed rapidly in the field of experimental data analysis and mining, emerging as a powerful new tool to address data mining challenges. Models such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) have been widely applied in image recognition, natural language processing, and other fields, achieving remarkable successes. However, to achieve accurate prediction results, deep learning models require a large amount of data for training and validation. The scale and quality of experimental datasets directly impact the training effectiveness and prediction accuracy of the model. Additionally, the distribution and diversity of test datasets affect a model's generalization ability and robustness. To ensure a model's generalization ability, it is crucial to simulate real-world scenarios as closely as possible and select representative data as the training and test sets. A key research area concerns the development of data enhancement techniques based on deep learning, which is significant for improving model performance, preventing overfitting, enhancing stability, and expanding the scope of application. Through the application of data enhancement technology, deep learning models can better adapt to various practical application scenarios, laying the foundation for their widespread application in various fields.