Supervised Learning: Predicting Movements in Social Security Filings

Summary:

Through the collection of data related to the US Financial System, I evaluate changes in the number of people receiving social security retirement benefits. Through utilizing different econometric and machine learning methodologies, I am able to successfully predict, with 92% F-1 score, if the number of social security recipients will rise or fall.

• Full writeup with code
• R Code

Models & Methods

• Logistic Regression
• Hypothesis Testing - Augmented Dicky Fuller
• Greedy Selection Methods (Exhaustive, Forward, Backward)

Performance

• F-1 Score: 92%
• Accuracy: 93%

Technologies

R