direct multi-step (DMS) forecasting

d-ailin / GDN

Implementation code for the paper "Graph Neural Network-Based Anomaly Detection in Multivariate Time Series" (AAAI 2021)

MIT License

482 stars 141 forks source link

direct multi-step (DMS) forecasting #104

Open kaya0007 opened 3 weeks ago

kaya0007 commented 3 weeks ago

Thank you for your excellent work! Is the aggregated feature z (t) obtained from formula (5) for a single time t, or does it include all past historical times? Is it theoretically feasible to directly map this feature to future multi-step outputs through MLP? How to explain interpretability? Thank you for your answer!