During the summer of 2018, I worked with Professor David Matteson at Cornell University and Professor Rebecca Killick at Lancaster University to implement a new R package: changepoint.online. Although the GSoC period ended in August, I continued developing the project and left an active GitHub repository for other researchers to extend. Further work, such as online non-parametric changepoint detectionm is now in progress at Lancaster University.

I used R, C++, and C to create the package while working from Lancaster University, gaining exposure to new statistical research beyond my degree curriculum.

The project moved beyond existing offline changepoint R packages by providing an online framework. It implements traditional “resetting” methodology (after a change, prior data is forgotten) and brings the accuracy benefits of offline methods to the online setting in a computationally efficient way.