pdprpr 0.7.2

PanDas PRePRocessor: Preprocess Pandas Objects for Machine Learning

PanDas PRePRocessor: Preprocess Pandas Objects for Machine Learning

Installation

$ pip install pdprpr

Usage

Assume you have following DataFrame to be preprocessed:

from pandas import DataFrame

df = DataFrame({
    'num': [1, 3, float('nan')],  # numerical feature, needs to be scaled in [0, 1]
    'cat': ['p', 'q', 'r'],       # categorical feature, needs to be transformted to dummy var
    'bin': [False, False, True],  # binary feature, needs to be 0 / 1
}, columns =['num', 'cat', 'bin'])
#    num cat    bin
# 0  1.0   p  False
# 1  3.0   q  False
# 2  NaN   r   True

Define preprocessing settings:

# preprocessing.yml
- name: num
  kind: numerical

- name: cat
  kind: categorical

- name: bin
  kind: binary

Then create DataFramePreprocessor with them:

import yaml

with open('preprocessing.yml') as f:
    settings = yaml.load(f)

from pdprpr import DataFramePreprocessor

processor = DataFramePreprocessor(settings)

Finally use it to preprocess the DataFrame:

processor.process(df)
#    num__VALUE  cat__p  cat__q  cat__r  bin__TRUE
# 0         0.0       1       0       0          0
# 1         1.0       0       1       0          0
# 2         NaN       0       0       1          1

Documentation

For more options please see tests untill docs get available…