In order to improve the effect of food security monitoring and early warning, this paper proposes a food security monitoring and early warning system combined with big data technology. Moreover, this paper constructs a food security combination prediction model based on machine learning to efficiently predict food production and marketing security, food financial security, and food insurance security, which helps to form a food security early warning model reserve pool and lays an early warning model foundation for intelligent decision-making. Then, this paper combines experiments to verify that the model proposed in this paper has more advantages in food security monitoring compared with the current popular deep learning models. At the same time, the experimental results show that the proposed method can effectively monitor food security problems in different situations, and can fuse high-dimensional sensor data for information fusion. In addition, the powerful self-attention mode can automatically learn the abnormal content in sensor data. Compared with the same type of anomaly detection models, this model has shorter training time and higher accuracy.