dst#1: Predictive modeling using R
Boris Cergol, Ektimo CEO: In this talk I will share some of the best things I have learned about building predictive models. I will show you that by using the right packages, R can turn into a highly useful machine intelligence tool. By presenting a case study about predicting future stock returns, I will guide you through an example workflow of a data scientist that includes steps such as:
– data preprocessing and normalization,
– generating and selecting the right features,
– splitting data into training, validation and test sets,
– learning a classification algorithm,
– fine tuning your model,
– evaluating how good the model actually is.
The case study involves a big-ish dataset so I will touch upon how you can be more efficient using parallel computation and which packages bail you out when your data grows to big for your memory. Working example code will be provided to all participants.