Hi Pros,
When we form our stock prediction question as a classification model to predict stock to go up or down in the future, the target variable Y can be set as N-day-ahead stock returns e.g. one-day-ahead stock returns or 30-day-ahead stock returns.
we set Y as one-day-ahead returns for a short-term prediction and set it as one-month-ahead returns for a long-term prediction.
So when it comes to yield a better prediction accuracy, should we optimize this parameter "N" by trying say 1 day, 5 days, 10 days, 15 days, 30 days..etc.. to see which value n yield the best accuracy?
Or is it not appropriate to have a comparison this way because it is not an apple to apple comparison?
Thank you for your answer in advance
Regards,
Kelvin