Extending the data into the past

ZhengyaoJiang / PGPortfolio

PGPortfolio: Policy Gradient Portfolio, the source code of "A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem"(https://arxiv.org/pdf/1706.10059.pdf).

GNU General Public License v3.0

1.74k stars 750 forks source link

Extending the data into the past #40

Closed ziofil closed 6 years ago

ziofil commented 6 years ago

Hi,

If a currency doesn't have enough data before a certain point in time, how do you extend it? With constant values? if so how do you pick the value?

I think I understood that you use the oldest non-NaN value, but some currencies have a crazy initial spike before settling very close to zero, so I'm not sure that's what you actually do.

dexhunter commented 6 years ago

possible #18 related

ziofil commented 6 years ago

Ooh, thank you, I had not seen that. Solved.