Closed tomespel closed 6 years ago
Some notes:
Am 05.03.2018 um 11:26 schrieb Mehdi Tomas notifications@github.com:
Some notes:
It is clever to split the dataset by date and asset as this will be useful in the next steps; I would also For feature engineering and manipulation, it would be more practical to merge volatility, returns, etc, in the form of a list of features instead of considering them separately. I need to think about it, but I'm not yet convinced that we need the values of each of those features in our KairosDay class. I think we could go further with your ideas and drop the localClassification element and encode everything within the classificationDetails, especially since I don't think there will be a need to access them in this format and it would simplify the class — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub, or mute the thread.
Replying to your second point, I will implement a new structure. All the values will be stored into a single vector (ˋlistˋ format), a Turing-like data ribbon. A dictionary will keep track of the location on the ribbon (the adresses as a pair of ˋintˋ with first value and length). This will increase the output time when extracting to a ML tool, while preserving the ability to keep track of the elements. I’ll try to implement this before tomorrow night, it shouldn’t be long.
Am 05.03.2018 um 11:26 schrieb Mehdi Tomas notifications@github.com:
Some notes:
It is clever to split the dataset by date and asset as this will be useful in the next steps; I would also For feature engineering and manipulation, it would be more practical to merge volatility, returns, etc, in the form of a list of features instead of considering them separately. I need to think about it, but I'm not yet convinced that we need the values of each of those features in our KairosDay class. I think we could go further with your ideas and drop the localClassification element and encode everything within the classificationDetails, especially since I don't think there will be a need to access them in this format and it would simplify the class — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub, or mute the thread.
Here are the elements of the class KairosDay.
__id (int)
corresponds to the target id.__asset (int)
corresponds to the asset id.__date (int)
corresponds to the date.__features (list)
is the continuous list of all features in the class.__featuresIndex (dict)
specifies the position of the feature in __features (list)
in the form of a dictionary {'featureName': (featureLocation (int), featureLength (int))}
.__localClassification (string)
specifies the local classification hlv, mlv or llv. States None otherwise.__classificationDetails (dict)
specifies the local classification probabilities in the form of a dictionary {'method': double(probability)}
.
Following #4 . Here are the elements of the class
KairosDay
.__id (int)
corresponds to the target id._asset (int)
corresponds to the asset id.__date (int)
corresponds to the date.__volatility
corresponds to the list ofdouble
with the volatility values.__returns
corresponds to the list ofdouble
with the returns values.__isClassified (bool)
indicates if the asset has received it local volatility classification.__localClassification (string)
specifies the local classificationhlv
,mlv
orllv
. StatesNone
otherwise.__classificationDetails (dict)
specifies the local classification probabilities in the form of a dictionary{'method': double(probability)}
.__features (dict)
specifies the features list in the form of a dictionary{'featureName': list(featureValuesList)}
.