Time Series Generation for Agricultural IoT Using TTS-GAN

With the rapid development of Internet of Things (IoT) technologies, the agricultural sector has gained significant technical advantages in efficiently collecting time-series data. However, in practical applications, data acquisition is often constrained by factors such as privacy regulations and unstable network connectivity, resulting in insufficient data volume, poor continuity, and suboptimal quality. To address these challenges, this study employs TTS-GAN, an innovative model that integrates Generative Adversarial Networks (GANs) with a Transformer-based architecture, to generate high-quality synthetic agricultural time-series data. The experimental evaluation, conducted on a single agricultural IoT dataset, demonstrates that TTS-GAN outperforms TimeGAN and RCGAN on several metrics. In addition, dimensionality reduction visualizations using PCA and t-SNE reveal a high degree of distributional overlap between real and generated data, underscoring the model’s ability to produce realistic and representative samples. This research offers an effective solution to the problem of data scarcity in agricultural time-series analysis and provides new technical pathways and methodological references for related studies.