Predictor insight graphs

Introduction to Predictive Analytics in Python

Nele Verbiest, Ph.D

Data Scientist @PythonPredictions

Motivation for predictor insight graphs

1. Build model
2. Evaluate model using AUC
3. Evaluate model using cumulative gains and lift curves
4. Verify whether the variables in the model are interpretable

Interpretation of predictor insight graphs

Predictor insight graphs for continuous variables

The predictor insight graph table

Income	Size	Incidence
low	20850	0.0431
average	62950	0.0492
high	16200	0.0615

 

print(pig_table["Size"][income=="low"])

20850

Constructing a predictor insight graph

• (Discretisation of variable if continuous)
• Calculate predictor insight graph table
• Plot the predictor insight graph

Let's practice!

Introduction to Predictive Analytics in Python