ADMM: Computational Analysis of Lasso and Ridge on Fraud Detection

Alternating Direction Method of Multipliers or ADMM is becoming increasingly vital in resolving regression and classification problems with complex datasets containing a variety of attributes or features. As datasets grow in complexity, ADMM's ability to efficiently handle large-scale optimization problems becomes crucial for achieving accurate and effective solutions in regression and classification tasks. Its versatility makes it well-suited for addressing the challenges posed by intricate datasets, ensuring optimal performance in various domains such as machine learning, signal processing, and image analysis. This paper compares the implementation of both LASSO and Ridge regression techniques using ADMM with the traditional non-distributed single-machine implementation of both techniques. Through simulations, the significantly reduced computational time achieved by the regression models optimised by ADMM is highlighted compared to the standard regression approaches.