ModelStacking

ModelStacking

The function creates a logistic regression model using a stacking scheme.

Parameters:

• models_dict: dict
  A dictionary with LogitModel instances.
• x_data: pd.DataFrame
  A pandas DataFrame with explanatory variables.
• y_data: pd.Series
  A pandas Series with discrete dependent variable.
• penalty: Optional[str], default = None
  A regularization for Logistic Regression model. It can be either 'None', 'l1', or 'l2'.
• alpha: float, default = 0.5
  A float variable for the regularization.
• fit_intercept: bool, default = True
  A boolean variable whether to fit intercept in the Logistic Regression or not.
• logprob: bool
  An indicator to calculate the logarithm of probabilities.

Returns:

• model: LogisticRegression
  A LogisticRegression model from the sklearn package.

Exceptions:

• ValueError:
  Raised if the length of x_data and y_data must be identical
  Raised if the penalty is not in [None, l1, l2]
  

• TypeError:
  Raised if models_dict is not a dictionary
  Raised if x_data is not a pandas DataFrame object
  Raised if y_data is not a pandas DataFrame object
  

Example:

from combat.combat import ModelStacking

# Sample input data
models_dict = {1: model1 # LogitModel instance
              , 2: model2 # LogitModel instance
              , 3: model3 # LogitModel instance
              }
x_data = your_data
y_data = your_labels

# Create a stacked logistic regression model
stacked_model = ModelStacking(models_dict, x_data, y_data, penalty='l1', alpha=0.5, fit_intercept=True, logprob=False)
print("Stacked model:", stacked_model)