ElasticNetFactorsIdentifier
- class qf_lib.common.utils.factorization.factors_identification.elastic_net_factors_identifier.ElasticNetFactorsIdentifier(max_number_of_regressors: int = 10, epsilon: float = 0.05, l1_ratio: float = 1, number_of_alphas: int = 75, is_intercept: bool = True, graphic_debug: bool = False)[source]
Bases:
FactorsIdentifier
Class used for identifying factors in the model with Elastic Net method (with Cross-validation).
- Parameters:
max_number_of_regressors – maximal number of regressors that can be included in the model
epsilon – if abs(coefficient) is smaller than epsilon it is considered to be zero, thus won’t be included in the model
l1_ratio – value between [0,1] the higher the simpler and more sensitive model is to collinear factors
number_of_alphas – number of different lambda values tested
is_intercept – True if intercept should be included in the model, False otherwise
graphic_debug – default False; If True, then some additional debug info will be plotted (used when tuning the ElasticNetFactorsIdentifier’s parameters)
Attributes:
minimal number of regressors taken for the alpha_1se (max.
number of folds in the k-fold cross-validation.
Methods:
select_best_factors
(regressors_df, analysed_tms)Returns the dataframe which is the subset of the original regressors_df but only contains rows for dates common for it and analysed_tms and only contains columns for coefficients which should be included in the model.
- MIN_NUM_OF_1SE_REGRESSORS = 2
minimal number of regressors taken for the alpha_1se (max. alpha for which the MSE is within 1 std. from the min. MSE). If number of regressors is smaller, then coefficients for min. MSE are taken.
- NUMBER_OF_FOLDS = 10
number of folds in the k-fold cross-validation.
- select_best_factors(regressors_df: QFDataFrame, analysed_tms: QFSeries) QFDataFrame [source]
Returns the dataframe which is the subset of the original regressors_df but only contains rows for dates common for it and analysed_tms and only contains columns for coefficients which should be included in the model. Factors are identified using Elastic Net method with Cross-Validation (for calculating the MSE).
- Parameters:
regressors_df – dataframe containing data for regressors (e.g. daily log-returns)
analysed_tms – timeseries of analysed data (data which should be modeled with regressors, e.g. daily log-returns)
- Returns:
Subset of the original regressors_df. Only contains rows corresponding to dates common for it and analysed_tms. Only contains columns corresponding to coefficients which should be included in the model
- Return type: