Shaving a Classifier with Occam’s Razor

I recently went through a Coursera course on Classification taught by Carlos Guestrin from U of Washington and thought it was excellent.   There was an interesting discussion on model overfitting that I thought I would share. In previous posts I discussed linear models with shrinkage parameters such us ridge and lasso regression models. Similar approach … More Shaving a Classifier with Occam’s Razor

Fraud Detection with Benford’s Law

Today’s post is about a phenomena observed in many “naturally” occurring data collections that can be exploited in accounting fraud detection.  This phenomena is referred to as Benford’s Law and is named after Frank Benford who popularised it.  The original discovery is actually credited to Simon Newcomb. The main idea is that leading digits in … More Fraud Detection with Benford’s Law

Classification Trees

In today’s post I wanted to describe classification trees. I will concentrate on Classification and Regression Tree (CART) algorithm. As I discuss the main features of this algorithm I will contrast it with other widely used methods to construct classification trees. What A Classification Tree Looks Like: Before diving into the algorithm let’s have a … More Classification Trees

Lasso model example (LME’s Aluminium Futures Price)

In my previous post I showed a coordinate descent algorithm for solving Lasso coefficients. Lasso model is part of a family of penalized regression models that are popular in machine learning and predictive modeling. In today’s post I want to show you how this model can be used to estimate the monthly average price of … More Lasso model example (LME’s Aluminium Futures Price)