Multi-task Learning-based Early Warning of Anomalous Risks in International Trade and Economic Data

The current international trade and economic data abnormal risk warning mechanism is generally fixed, and for the target-based one-way warning processing, covering a small scope of early warning, resulting in the ultimate risk of early warning to capture the frequency of decline, so that the proposed multi-task learning based on the international trade and economic data abnormal risk warning method design and analysis. According to the current determination, the first abnormal data identification and preprocessing, the risk type synchronization determination, the use of multi-level approach, expand the coverage of the warning range, set the multi-level warning mechanism. On this basis, we constructed a multi-task learning data anomaly risk early warning model for economic and trade data, and realized the anomaly risk early warning with continuous monitoring assistance. The test results show that compared with the visualization of abnormal data early warning processing mode, gray fuzzy comprehensive evaluation data early warning method, this design of multi-task learning of international trade and economic data abnormal risk warning method finally came up with the risk of early warning to capture a relatively high frequency, which indicates that the design of the international trade and economic data abnormal risk early warning method is more efficient, accurate, stable, early warning processing effect is significantly improved, with a practical application value.