This is neural net trading application.
Now development but 1st iteration of training is already done.
Now development feature of #1.
This is not a library but customizable easily via add the pipeline code like this.
src/pipelines.py
from pipeline import PipeLine
def train_pipeline():
saved_model_path = "models/saved_model.h5"
data_path = 'data/bitstampUSD_1-min_data_2012-01-01_to_2019-03-13.csv',
# pipeline definition
PipeLine(
data_path,
ReadData(),
PreprocessingProcedure1D(),
TrainProcedureKeras(
KerasLinear1D(
saved_model_path=saved_model_path
)
)
).execute()pipeline is defined at src/pipeline.py
PipeLine constructor never take argument which are not extend Procedure except 1st argument.
Each procedure are implement Procedure class.
So you can add procedure easily.