A GANs Based Data Synthetic Technique for Enhancement of Prediction Accuracy of Mushroom Classification

Abstract

In the landscape of machine learning and data-driven decision-making, limited data availability often undermines classification model accuracy. This study pioneers a solution by leveraging Copula Generative Adversarial Networks (Copula GANs) to generate high-fidelity synthetic data, with a focus on mushroom classification. Copula GANs replicate original dataset characteristics, as confirmed by thorough Category Coverage and TV Complement evaluations that validate its ability to accurately emulate category distributions and multivariate dependencies. To substantiate the practical impact, a mushroom classification task employs a decision tree. Results showcase notable accuracy enhancement through the integration of synthetic data with real data. Fine-tuning Copula GAN parameters, exploring feature interpretability, extending the technique to diverse domains, and merging with traditional data augmentation methods are promising future avenues. In essence, this study pioneers Copula GAN-generated synthetic data as a novel solution to data scarcity. The outcomes highlight the efficacy of synthetic data augmentation, advancing the potential of machine learning models across real-world applications.