Gumly

Features

- module geo_location.py
- state_to_uf: Fill the region through the state returning a pd.series.
- uf_to_state: Fill the region through the state returning a pd.series.

- module imbalanced.py
- oversampler: Runs the chosen method of Over-sampling at imbalanced data
and returns the balanced tuple with two arrays at index 0
the values of input "X" transformed and at index 1 the values of "y".
- undersampler: Runs the chosen method of Under-sampling at imbalanced data
and returns the balanced tuple with two arrays at index 0
the values of input "X" transformed and at index 1 the values of "y".
- combine: Runs the chosen method Combination of over-and undersampling at imbalanced data
and returns the balanced tuple with two arrays at index 0
the values of input "X" transformed and at index 1 the values of "y".

- module value_validation.py
- check_int: This function returns True if the first argument is an integer and it
lies between the next two arguments. Otherwise, returns False.
- assert_check_int: This function uses the check_int function and it will raise an exception if it returns False.
- check_list: Checks whether the given value is a list. Optionally, verifies if it has
some specific number of elements and/or if all elements are instances of a
given type/class. It returns True if all checks pass and False otherwise.
- assert_check_list: This function uses the check_list function and it will raise an exception if it returns False.

- module checkpoint_flow.py
- LocalStateHandler: his class is responsible for writing and reading the contents of the state of
a program (i.e. a set of small variables) into and from a file in the local
filesystem. The state is defined as a dictionary and it is serialized in
the file with the Dill library.
- CheckpointFlow: This class is responsible for creating and running a series of functions
while checkpoint along the way, so that if something goes wrong in any
function, when re-running the series, it can start from the function which
raised the error, instead of running everything all over again.

Features

- module geo_location.py
- city_to_region: Fill the region through the city returning a pd.series.
- city_to_microregion: Fill the microregion through the city returning a pd.series.
- city_to_mesoregion: Fill the mesoregion through the city returning a pd.series.
- city_to_immediate_region: Fill the immediate region through the city returning a pd.series.
- city_to_intermediary_region: Fill the intermediary region through the city returning a pd.series.
- state_to_region: Fill the region through the state returning a pd.series.
- cep_to_state: Fill the state through the cep returning a pd.series.
- cep_to_region: Fill the region through the cep returning a pd.series.
- ibge_to_city: Fill the city through ibge id returning a pd.series.
- city_to_ibge: Fill the ibge id through the city returning a pd.series.

Features

- module dimencionality_reduction.py
- dimencionality_reduction: Wrapper for two well-know methods, PCA and svds.

- module feature_engineering.py
- split_features_and_target: Separates the features and the target columns into two new dataframes.
- feature_selection_filter: Feature selection using filter technique and chi2 values.
- feature_selection_wrapper: Feature selection using wrapper technique and LogisticRegression.
- feature_selection_embedded: Feature selection using embedded technique and LightGBMClassifier.
- feature_selection_stepwise: Perform a forward-backward feature selection based on p-value from statsmodels.api.OLS.
- feature_selection_f_regression: Perform a f_regression feature selection based on p-value.
- feature_selection_mutual_information: Perform a mutual_info_regression feature selection.
- ordering_filter: Analyzes the records of all given variables and returns their indexes.

- module files.py
- create_dir: Creates a directory if it doesn't exist.
- move_files: Moves the files from source_dir to dest_dir.

- module hyperparameter_tuning.py
- hyperparameter_tuning: Perform hyperparameters optimization using optuna framework for the chosen technique.

- module metrics.py
- weighted_mean_absolute_percentage_error: Implements the weighted MAPE metric.

- module value_validation.py
- check_number: This function returns True if the first argument is a number and it lies between the next two arguments. Othwerise, returns False.
- assert_check_number: This function uses check_number function and it will raise an exception if returns False.
- check_dtypes: Verify the list of types.
- assert_check_dtypes: Verify dataframe columns dtypes.