implementation of WSAE-LSTM model as defined by Bao, Yue, Rao (2017)
Folder structure updated:
Report output from visualize.py
in reports
folder:
Sample of report output from visualize.py
:
scaling and denoising moved into separate files in features
folder:
scaled data and scaled + denoised data stored in 'data/interim` folder:
-thresholding mode changed from soft to hard to prevent division by 0/error in true_divide for numpy
data/pickled
folder for pickled data (i.e. pickled dataframes)references
README.md
update to reflect changesutils.py
: interval_split(), dict_interval_split(), pickle_save, pickle_load()
(refactored from 1d_train_test_split_exploration.ipynb
)requirements.txt/environment.yml
):
from statsmodels.robust import mad
from scipy import signal
1c_wavelet_draft_test_exploration.ipynb
1f_tvt_split_exploration.ipynb
interval_split()
in utils.py
wsae_lstm
folder to clean raw dataset, output stored in data/interim
folder (refactored from notebooks\0_data_clean_load.ipynb
notebook)1_data_clean_load_datetime.ipynb
notebook:
utils.py
for frames_to_excel(),dictmap_load(), dictmap_datetime()
functions in wsae_lstm
source/root folder
clean_data.py
updated with function imports from utils.py
frames_to_excel()
can now accept optional key_order
kwargdata/processed
has date column in datetime object formatreadme.md
update with repository structure section & other minor clarification changes