MetaSieve

Implementation of MetaSieve algorithm. (For "MetaSieve: Performance vs. Complexity Sieve for Time Series Forecasting" (OEDM workshop of the IEEE International Conference on Data Mining (ICDM'22)).

predictions

• LSTM_prediction.py - Generating synthetic data. Implementing brute-force calculations of quality metric for all generated sequences.
• LSTM_prediction.py, RF_prediction.py, and XGB_prediction.py - files which contains code for obtaining predictions accuracy for 15 levels of LSTM, RF, and XGBoost models respectivly.

results

• acc_time.ipynb - Notebook, which provides the research of time efficiency of different strategies of MetaSieve.
• seive_drawing.ipynb - Notebook, which provides the research of MetaSieve results with the usage of different quality control strategies.
• GNNclass.ipynb - realization of GNN classifier for the Second sttage of MetaSieve.

storage

• artdata_1000.csv - generated 1000 synthetic time-series.
• real_data.csv - real-world data consisting of stock value and electric consuption time-series.
• art1000_LSTM_acc_time.csv, art1000_RF_acc_time.csv, art1000_XGB_acc_time.csv - brute-forse sMAPE and RMSE results for synthetic data with measured time.

other

• ArtComposer.py - generation process of 1000 synthetic time-series.