If you start building an open source application for data mining, what algorithm would you like to see implemented first for data analysis? And why?