The invention provides a chronic disease aggravation risk assessment and alarm system. The system comprises a preprocessing module, a core index feature extraction module, a multi-dimensional time sequence data feature extraction module and a feature fusion and early warning module. The preprocessing module carries out comprehensive preprocessing on medical data to obtain a historical sequence ofcore index data and the historical data of multi-dimensional features. The core index feature extraction module establishes a correlation between a time interval between adjacent time points in a historical sequence of the core index data and long-term memory transfer to obtain a first group of reference features. The multi-dimensional time sequence data feature extraction module extracts the multi-dimensional time sequence data to obtain a second group of reference features. The feature fusion and early warning module fuses the first group of reference features and the second group of reference features, and then performs early warning analysis.