Anomaly Detection for Internet of Things Time Series Data Using Generative Adversarial Networks with Attention Mechanism in Smart Agriculture

In recent time, smart agriculture which is the integration of modern agriculture and Internet of Things (IoT) technology has received widespread attention and developmental process. In order to actualize scientific cultivation and precise control, the agricultural environment is monitored in real time by deploying various relevant sensors. Large amount of multidimensional time series data is generated. However, smart agricultural IoT architectures often suffer from data loss and misrepresentation. Moreover, some intelligent decision-making approach for agricultural management also requires detailed data analysis. To address the aforementioned problems, this paper developed a new anomaly detection model based on generative adversarial networks (GAN), to process the multidimensional time series data generated. GAN is a deep learning model trained to learn the distribution patterns of normal data and capture the temporal dependence of time series and the potential correlations between features through learning. To address the issue of generator inversion, an encoder–decoder structure incorporating the attention mechanism is designed to improve the performance of the model in learning normal data. In addition, a new reconstruction error calculation method that measures the error in terms of point-wise difference and curve similarity to improve the detection effect were presented. Finally, considering three smart agriculture-related datasets, experimental results showed that the developed model accurately achieved anomaly detection. The experimental precision, recall, and F1 score exceeded the reference models with 0.9351, 0.9625, and 0.9482, respectively.