Gather remaining evidence

[LiFaytheGoblin]

• create test and training sets and return them
• train model
• return trained model as .ser
• get predictions and return them

Known issues:

• sampleid as currently stored in CSV is not always unique, because second part (timesplitting) was removed. Needs fixing.