This repo is the origin pytorch implementation of MIANet in the following paper: MIANet: Multi-level temporal information aggregation in mixed-periodicity time series forecasting tasks.
- The code is based on the LSTNet. We've mainly modified the models part of it, and the rest has only made simple adjustments;
- You can download the Traffic, Electricity, Solar-AL datasets mentioned in the paper at https://github.com/laiguokun/multivariate-time-series-data;
- For Ausgrid_GC, Ausgrid_GG datasets, we provide the converted data at dataset. You can the raw data at https://forecastingdata.org;
- We provide a trained model of Traffic at horizon=12, and you can test it directly.