Skip to content

Niranjankumar-c/CreditRiskAnalytics

Repository files navigation

CreditRiskAnalytics

Business Context: Credit scoring is the set of decision models and their underlying techniques that aid lenders in the granting of consumer credit. Typical application areas in the consumer market include: creditcards, autoloans, home loans and a wide variety of personal loan products.

Objective

  • In this case study, One of the leading banks would like to predict customers who are most likely to default on the loan.
  • For new customers we need to decide whether to extend credit or not by analyzing the behaviour of existing customers.

Project Summary

Project Summary

  • ROC Curve - An ROC curve (receiver operating characteristic curve) is a graph showing the performance of a classification model at all classification thresholds.
  • Finding the best cufoff point - which maximises the specificity and the sensitivity. This threshold point might not give the highest prediction in your model, but it wouldn't be biased towards positives or negatives.

ROC Curve

Cutoff Curve

Installation

The Installation process will get you a copy of the project up and running on your local machine for development and testing purposes

  1. Clone or download the project into your local machine.
  2. Unzip the project folder.
  3. Open the source file CreditRiskAnalytics-DefaultModel using JypyterNotebook and execute the file.
  4. Instead of step 3, use the classification model loaded in the pickle to classify default customers.
import pickle
# load the model from disk - use to classify the default customers directly
loaded_model = pickle.load(open('OutPutModel/final_model.pkl', 'rb'))
print("Loaded Decision tree model :: ", loaded_model)
  1. After successfully loading the saved model, we can use them in the general way to predict for test dataset or in the production servers.

Prerequisites

The following list summarizes the packages/softwares used in this project. These are the softwares/packages you neeed to install before executing the project file.

  • Anaconda v – 5.2.0 (py 36_3)
  • Python v – 3.6.5
  • Packages (Packages that are not part of anaconda distribution, need to install using pip installer)

Repository Contains

License & CopyRight

Copyright (c) 2018 Niranjan Kumar Licensed under the Apache License 2.0

Trademarks

All other trademarks referenced herein are the property of their respective owners.