I am using XGBoost to try to predict the direction of the stock market based on social media sentiment. Having read through some studies, I was planning to separate the training/test data by time period, e.g. use 2014-2016 data for training and 2016-2018 data for testing.
Does that make intuitive sense given the nature of the data I am using?
I am happy to provide any further details which would be helpful, thank you.