Financial Applications of Machine Learning

My machine learning industry projects throughout the course in my Financial Engineering Program

• Built Gradient Boosting, Neural Network and Logistic Regression models after processing data and feature selection.
• Interpreted models to eliminate black box bias
• Ran grid search to fine tune the model's parameters and analyzed models stability (variance and bias).
• Sorted customers based on the probability of default estimated by each model, and segmented customers in to 10 bad rate buckets.
• Priced each buckets relevant interest rate and expected Net Present Value per segment.

Models Evalustion for Ordinal Bad Rates Score Ranking:

Logistic Regression Models Interest Pricing per Bucket:

Projected Revenue (Business) Model for Booking all Segments:

Build Lookalike Logistic Regression Model (Keras and SKLearn)

While tuning the neural netwrok model for loan prediction for later use to segment customer in to bad rate buckets, I thought of how a preceptron with sigmoid activation function and SKLearn's Logistic regression model can give same/approximately lookalike performance by understanding common hyperparameters.

For full metholodgy and brain storming discussion article!

Methods for Addressing Black Box Bias:

Interpretation Techniques