Question

Explain the easiest way to create credit risk algorithm to sort a list of loan applicants?

Answer 1

When the customer apply for a loan granting application, one must investigate about what called 5 C’s which are Character (or Credit History), Cash Flow (or Capacity), Collateral, Capitalization and Conditions. It is helpful for evaluation loan application and it regarded as a helpful framework for estimate the credit risk related to a probable creditor.

1) Dataset: After doing preprocessing tasks and applying feature selection to choose the most important attribute we used 8 attributes Credithistory, Purpose, Gender, Credit amount, Age, Housing, Job and the Class:

2) MODEL IMPLEMENTATION: Prepare the training set, a records that are already have known class label. Build the model by applying one of learning algorithm using training set. Applied the model upon unknown data test set class. Evaluate the accuracy of the model.

2.1) J48 classification algorithm: J48 is the enhanced edition of C4.5 algorithms or can be viewed as C4.5 implementation.J48 takes as an input the set of tables and generate a decision tree as an output. The generated decision tree is alike to the structure of tree. It consists of root, intermediate and leaf node. The nodes in the generated tree contain a decision which guide to the result. It split the input data set into mutually exclusive sets, each set with a label. Splitting measure is applied to determine which attribute lead to the optimal splitting

2.2) BayseNet algorithm: This algorithm depends on the theorem of Bayse. We build Bayesian Network after calculating condition probability to all nodes. It represent a directed acyclic graph.

2.3) NaiveBayes The Bayesian is a supervised learning method.It characterized with it is elegance, simplicity, and robustness. For this reason it is became widely used in classification purposes. it supposes that attributes of a class are self-determining in real life.

Therefore, J48, BayesNet and NaiveBayes algorithms can be used to build a predictive models that can be used to predict and classify the applications of loans that introduced by the customers to good or bad loan by investigate customer behaviors and previous pay back credit.