Learn R Programming

scorecardModelUtils (version 0.0.1.0)

Credit Scorecard Modelling Utils

Description

Provides infrastructure functionalities such as missing value treatment, information value calculation, GINI calculation etc. which are used for developing a traditional credit scorecard as well as a machine learning based model. The functionalities defined are standard steps for any credit underwriting scorecard development, extensively used in financial domain.

Copy Link

Version

Install

install.packages('scorecardModelUtils')

Monthly Downloads

151

Version

0.0.1.0

License

GPL-2 | GPL-3

Maintainer

Arya Poddar

Last Published

April 14th, 2019

Functions in scorecardModelUtils (0.0.1.0)

Recursive Decision Tree partitioning with monotonic event rate along with IV table for individual numerical variable

Computes error measures between observed and predicted values

Calculating mode value of a vector

Scoring a dataset with class based on a scalling logic to arrive at final score

Converting coefficients of logistic regression into scores for scorecard building

Univariate analysis of variables

support_vector_parameters

Hyperparameter optimisation or parameter tuning for Suppert Vector Machine by grid search

Missing value imputation

WOE and IV table for list of numerical and categorical variables

Binning numerical variables based on cuts from IV table

Redefines target value

gradient_boosting_parameters

Hyperparameter optimisation or parameter tuning for Gradient Boosting Regression Modelling by grid search

Clubbing of classes of categorical variable with low population percentage into one class

Removing multicollinearity from a model using vif test

Performance measure table with Gini coefficient, KS-statistics and Gini lift curve

random_forest_parameters

Hyperparameter optimisation or parameter tuning for Random Forest by grid search

Random sampling of data into train and test

Variable reduction based on Information Value filter

Clubbing class of a categorical variable with low population percentage with another class of similar event rate

Variable reduction based on Cramer's V filter

Clubbing class of categorical variables with low population percentage with another class of similar event rate

IV table for individual categorical variable

Creates confusion matrix and its related measures

Creates random index for k-fold cross validation

Pairwise Cramer's V among a list of categorical variables

Cramer's V value between two categorical variables

dtree_split_val

Getting the split value for terminal nodes from decision tree