In consideration of the current status of Rocket launch data interpretation, fault trend prediction and measures taken in advance would be benefit from studying of forecasting algorithms for the launch data. Rockets launch data is analyzed and then classified based on time series features. Aiming at the period-like data, the corresponding feature extraction algorithm is developed and the feature time series are reached. The rolling auto regressive algorithm is applied, with weighting history true value and forecasted value as the new modeling data at the present moment. The result is that the trend forecasting for the feature of the period-like data is realized, which will help to improve the rocket launch data interpretation.