Built predictive model for prediction of Probability of Default followed by calculation of Credit Score.
Considered multiple loans types across multiple banks for 5 years under the Central bureau and used a logistic regression.

Tools: R, SQL, Python
Techniques: GLM, CART, PCA, IV, WOE, GBM, RF
																	

																		Used AWS and machine learning techniques to build an ensemble of predictive models for estimating the probability of a credit card transaction being a fraud. Utilized big data of around 2 billion transactions with a very rare event rate for the exercise by segmentation and feature selection.

Highlights:
. Wrote an automated code for the whole process of model building starting from data pulling to metrics evaluation, inclusive of Sample Creation, Data Cleaning, Feature Selection and Dimension Reduction, Model Selection.
. Scaled up multiple processes and created a flow on high scale to match the size of the data along side optimizing the cost spent
. Worked in a highly time constrained environment managing the deadlines

Tools: R, Python, EC2, RedShift, S3
Techniques: GLM, CART, ANN, PCA, IV, WOE, GBM, RF, Firth Regression, Penalized Pregression