Multi-source time sequence data cleaning and time sequence matching method

The invention discloses a multi-source time sequence data cleaning and time sequence matching method, which comprises an initialization step, a cleaning step, a navigation body trajectory screening step and a navigation body trajectory data time sequence matching step, and is characterized in that a restoration candidate result is generated by adopting an autoregression model, and a cleaning result is evaluated by utilizing reinforcement learning; time sequence data points containing continuous errors are repaired, it is guaranteed that the cleaning result of minimum modification is always very close to a true value, then the optimal cleaning result closest to the true value is obtained in real time in the cross-domain multi-source environment, and therefore the data quality is improved; then, through a time sequence matching mode, the heterogeneous multi-source test data time sequence is standardized and unified, and available data support is provided for multi-domain joint analysis and evaluation. 本发明公开了一种多