- Number of videos:
We show how to write your own robust linear estimator within the Scikit-Learn framework using as an example the Theil-Sen estimator known as "the most popular nonparametric technique for estimating a linear trend".
The talk gives a small introduction of how Blue Yonder applies machine learning and Predictive Analytics in various fields as well as the challenges of Big Data. Using the example of Blue Yonder's machine learning software NeuroBayes, we show the made efforts and hit dead ends in order to provide a flexible and yet easy to use interface for NeuroBayes to Data Scientists. Since NeuroBayes is written in FORTRAN for performance reasons different interface approaches were tried which lead us eventually to a Python interface. In the talk we elaborate on the up- and downsides of the different approaches and the various reasons why Python won the race with an emphasize on the benefits of the Python ecosystem itself. Also, we discuss performance as well as scalability issues with Python and how we address them. In detail, we show the application of Cython to speed up calculations in the Python interface layer as well as distributed computing in a private cloud called Stratosphere. Scalability and efficiency is of utmost importance when data processing is time critical. Our overall goal is to give the audience an overview how Python fits in the software ecosystem of a company handling Big Data.