uyanik/shapka

An open source library for key drivers analysis based on Shapley values and Kano Theory

ShapKa: Customer Satisfaction Key Drivers based on Shapley values and Kano model

Installation

Use the following command to install the package:

pip install ShapKa

Usage

Use the following command for a key dissatisfaction drivers analysis (kda) :

import pandas as pd
from ShapKa.kanomodel import KanoModel

# Load data
df = pd.read_csv('data/example_03.csv')

# Define X and Y variables names
y_varname = 'Overall Satisfaction'
weight_varname = 'Weight'
X_varnames = df.columns.values.tolist()
X_varnames.remove(y_varname)
X_varnames.remove(weight_varname)

# Run analysis to identify key dissatisfiers
model = KanoModel(df, 
                  y_varname, X_varnames, 
                  analysis = 'kda',
                  y_dissat_upperbound = 6, y_sat_lowerbound = 9,
                  X_dissat_upperbound = 6, X_sat_lowerbound = 9,
                  weight_varname = weight_varname)

kda = model.key_drivers() ;kda

Here is the ouput :

Replace 'kda' by 'kea' in the analysis parameter if you want to identify key enhancers (kea) instead of key dissatisfiers

Documentation

• Documentation: https://shapka.readthedocs.io.

Credits

• The ShapKa package is based on the methodology developped by W. Michael Conklin, Ken Powaga and Stan Lipovetsky
• Some parts of the code are based on functions implemented in the Open Source Sage Mathematical Software

References

• Conklin, Michael & Powaga, Ken & Lipovetsky, Stan. (2004). Customer satisfaction analysis: Identification of key drivers. European Journal of Operational Research. 154. 819-827. 10.1016/S0377-2217(02)00877-9.
• Sage - Open Source Mathematical Software : https://github.com/sagemath/sage

shapka's Issues

Weighting Data

Is there a possibility to use a weight in your module like in a other learners (e.g. linear regression).

import numpy as np
import pandas as pd  
from sklearn.linear_model import LinearRegression
from sklearn.datasets import load_boston
import shap


boston = load_boston()
regr = pd.DataFrame(boston.data)
regr.columns = boston.feature_names
regr['MEDV'] = boston.target

X = regr.drop('MEDV', axis = 1)
Y = regr['MEDV']
w = weight # fictive sample weight

fit = LinearRegression().fit(X, Y[, weight])