adrianantico/retrofit

Automation of feature engineering, machine learning, model evaluation, model interpretation, eda, forecasting, recommender systems and more.

retrofit's Issues

ML Models: load data via numpy instead of pandas

Currently I'm using data conversion to Pandas frames to create model data but I want to switch it to numpy conversion as it's faster

Create plotting helper functions

Create Master Class for Plotting and subclasses for plotting library used. Plotting libraries should include:

• plotly
• matplotlib
• seaborn

Plotting functions should include:

• ScatterPlot()
• LinePlot()
• BarPlot()
• Histogram()
• 3D_ScatterPlot()
• HeatMap()

AutoDiff needs to be built for Polars and Pandas

The FE0_AutoDiff function uses imports from datatable to manipulate the data but these imports do not work on polars dataframes. I think it is missing a conditional processing block that will process a dataframe using polars' methods.

>>> areas.to_list()
['FLC', 'WAW', 'UTN', 'LAR', 'LAX', 'PHX', 'MIE', 'ATL', 'DFN', 'ONT']
>>> new_data = fe.FE0_AutoDiff(data, DateColumnName="dates", DiffNumericVariables=areas.to_list(), InputFrame="polars", OutputFrame="polars", Processing="polars")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\ProgramData\Miniconda3\lib\site-packages\retrofit\FeatureEngineering.py", line 617, in FE0_AutoDiff
    data = data[:, f[:].extend({Ref2: dt.shift(f[rcn], n = NLag2)})]
UnboundLocalError: local variable 'f' referenced before assignment

Save ml models

Needs to set up a function to save selected models to file for later use

ML Load Models

Need to set up a function to load models for scoring purposes

Add ML Evaluation Visuals

Depends on target type, but replicate existing plots from RemixAutoML