Linear Regression don't buy anything

[aateg]

The model learned that selling is the best decision for the scenario I'm testing.

Since we just allow to sell when the portfolio contains some stock to sell, the decision in the simulation is always holding.

Possible solution options:

• Test model in the different dataset (as suggested by Agoston)
• Improve the training strategy
  • Change the imputation method
  • Cross-validation (?)
• Allow negative portfolio (short selling)

In any case, we need to have some method for evaluating the models, the simulation is not sufficient to judge as something is good or not.

[aateg]

@agostontorok do you think we should allow short selling portfolio?