With ahead-of-time aircraft management, we are able to re- duce aircraft collision and improve air traffic capacity. How- ever, there are various impact factors which will cause a large deviation between the actual flight and the original flight plan. Such uncertainty will result in an inappropriate decision for flight management. In order to solve this problem, most of the existing research attempt to build up a stochastic trajec- tory prediction model to capture the influence of the weather. However, the complexity of the weather information and vari- ous human factors make it hard to build up an accurate trajec- tory prediction framework. Our approach considers the prob- lem of trajectory deviation as the ‘‘anomaly’’ and builds up an analytics pipeline for anomaly detection, anomaly diagnos- tics, and anomaly prediction. For anomaly detection, we pro- pose to apply the CUSUM chart to detect the abnormal tra- jectory point which differs from the flight plan. For anomaly diagnostics, we would like to link the entire anomalous trajec- tory sequences with the convective weather data and extract important features based on time-series feature engineering. Furthermore, XGBoost was applied to detect the anomalous trajectory sequences based on the time-series features. For anomaly prediction, we will build up a point-wise prediction framework based on the Hidden Markov Model and Convectional LSTM to predict the probability that the pilot would deviate from the flight plan. Finally, we demonstrate the sig- nificance of the proposed method using real flight data from JFK to LAX.