Closed XFeiF closed 3 days ago
Both the market information and the 158 factors are considered as features, which means they undergo the same preprocessing. Sorry that I missed pointing this out earlier, and please refer to Readme/Dataset/Preprocessing
for a description.
Specifically, the range of market information is [-3,3]
, which is due to the RobustZScoreNorm
in Qlib. It is a common practice in Qlib and please refer to the Qlib documents for explanation. You may also find my answer to this closed issue helpful.
Based on the Qlib, a version of data preprocessing code is already available, which includes detailed code to calculate market information from raw data. Please refer to Readme/[New]A-Qlib-implementation
.
feature.1
can not be inferred from the columnfeature
.For example, at
2008/1/8
, thefeature.1
's value is 1.0515707. While according to the expressionMask(Mean($close/Ref($close,1)-1,5),'SH000300')
, I think it can be simply calculated by(0.74735785 + 0.57677513 + 0.9783356 + 1.1510324 -0.4920946 ) / 5
, and the result is 0.592281276.Do I misunderstand the expression of
feature.1
? Can anyone explain this? Besides, can you provide the detailed code about how Market information is calculated?Many thanks!