The goal of this project is to implement some Machine Learning Algorithm in order to manage a feature selection process in a walk forward simulation.
In the project there are 400 features, 8 different time series and 1 Million time steps in each time series.
For each time step you have:
- A readout for each of the 400 features,
- A performance measurement,
Your algorithm needs to train, test and select the feature’s set that would most likely maximize the future performance of the system in some time steps ahead. You need to use the walk forward simulation using a sliding window not larger than 50% of the entire data set.
I expect you to deliver a C++ code to work with a VS2012 Compiler under Win 7 64x. You may use external libraries or packages subject to my prior consent.
We may arrange for you to use a third party HPC or cloud to work on this project in case you lack the necessary computing power.
I would need to agree with you on the acceptance criteria as regards to the efficiency of your methodology after you familiarize with the data but prior to awarding of the project to you.