solver_ops#

Solver generic methods

wip.modules.solver_ops.adjust_real_cost(real_cost, features, mult_coef=1, div_coef=3)[source]#

Adjust the real cost of some features

wip.modules.solver_ops.custom_format(x)[source]#

Format numeric value as a 5 decimal places string.

If :param:`x` is not a numeric value, the function returns the original value unchanged.

Parameters

x (Any) – The value to try to format as a string with 5 decimal places.

Returns

The formatted value or the original value if :param:`x` is not a numeric value.

Return type

Any

Examples

>>> custom_format(1.2345678)
'1.23457'
>>> custom_format('1.2345678')
'1.2345678'
>>> custom_format(0)
'0.00000'
>>> custom_format('some text')
'some text'
wip.modules.solver_ops.define_range_constraints(token: str, range_start: int, range_end: int, step: int = 1) → List[str][source]#

Generate a list of strings by applying a token format over a defined range.

This function receives a formatting token, and applies this format to each number in the range specified by range_start, range_end, and step. The formatted strings are then returned in a list.

Parameters

  • token (str) – A string with {} as a placeholder for the integer to be formatted.

  • range_start (int) – The start of the range to which the token is applied.

  • range_end (int) – The end of the range to which the token is applied. This value is not included in the output list.

  • step (int, optional) – The step between consecutive integers in the range, default is 1.

Returns

A list of strings obtained by applying the token to each number in the specified range.

Return type

list

Examples

>>> token = "{}_formatted"
>>> define_range_constraints(token, 1, 4)
['1_formatted', '2_formatted', '3_formatted']

>>> token = "prefix_{}_suffix"
>>> define_range_constraints(token, 1, 5, 2)
['prefix_1_suffix', 'prefix_3_suffix']
wip.modules.solver_ops.get_pi_system_tag_names(dataset: DataFrame) → DataFrame[source]#

Get the PI System tag names from the PIMS tag names.

This function takes a DataFrame and returns a copy of it with a new column containing the PI System tag names.

Parameters

dataset (pd.DataFrame) – pandas.DataFrame containing the optimization model resultas, with PIMS tag names as a column named ‘Tag’.

Returns

A pandas.DataFrame with the PI System tag names in a column named ‘Tag PI’.

Return type

pd.DataFrame

wip.modules.solver_ops.retrieve_best_model(model, models_results, metric='mape')[source]#

Sort the models by its metrics

wip.modules.solver_ops.retrieve_model_coeficients(model: str, models_results: dict)[source]#

Retrieve the ridge regression model coefficients.

Parameters

  • model (str) – Name of the model.

  • models_results (dict) – Dictionary with the models’ results.

Returns

Coefficient of each feature used in the model

Return type

zip

wip.modules.solver_ops.save_solver_results(solver_path, df, resultado_otimizador_filename: str = 'resultado_otimizador-US08_2024-04-19.xlsx')[source]#

Save optimization results to Azure Container Storage, or to a local filepath.

The default file name that is used to save the optimization results to Azure Data Lake is:

f"resultado_otimizador-US{US_SUFIX}_{datetime.today().strftime('%Y-%m-%d')}.csv"

For example, the file name should be similar to the following:

"resultado_otimizador-US08_2024-03-21.csv"

Parameters

  • solver_path (str) – The Azure Data Lake container URL path or a local folder path where the optimization results will be saved.

  • df (pd.DataFrame) – A pandas.DataFrame containing the optimization results for all production ranges.

  • resultado_otimizador_filename (str, default wip.constants.RESULTADO_OTIMIZADOR_FILENAME) – The name of the file to use for saving the optimization results.

  • versionchanged: (..) – 2.8.11: Bugfix error that caused optimization results to try to save it as a “.csv” file using the suffix “.xlsx” on the filename. This problem was making the DataBricks job “Integração SensUP” to read the optimization results in an incorrect format and subsequently fail to finish executing the job.

wip.modules.solver_ops.unnormalize_optimization_tags(scalers: dict, real_cost: dict) → dict[source]#

Unnormalize optimization tags using scalers.

This function receives scalers and real cost dictionaries, filtering the optimization keys, and unnormalizes the real cost based on the provided scalers. The real_cost dictionary is updated for each optimization key by dividing the value of the key by its corresponding data range in scalers.

Parameters

  • scalers (dict) – Dictionary containing the scaler objects, where each key represents a column, and the corresponding value is the scaler object associated with that column.

  • real_cost (dict) – Dictionary containing real cost data, where each key is a column name, and the corresponding value is the real cost associated with that column.

Returns

real_cost – Updated dictionary containing the unnormalized real cost data.

Return type

dict

Notes

This function assumes that the scalers dictionary contains Scaler objects with a data_range_ attribute. The function does not return a new dictionary object, rather it updates the input real_cost dictionary in-place and returns it.

The keys in real_cost that are in constants.TARGETS_IN_MODEL.values() or not in scalers.keys() are not considered as optimization keys, and thus are not processed.

Examples

>>> from sklearn.preprocessing import MinMaxScaler
>>> scalers = {'key1': MinMaxScaler(data_range=(1, 10))}
>>> real_cost = {'key1': 5, 'key2': 10}
>>> unnormalize_optimization_tags(scalers, real_cost)
{'key1': 0.5, 'key2': 10}
wip.modules.solver_ops.write_constraint(file, constraint, terms, target=False, description=False)[source]#

Write a set of constraints to a file.

Parameters

  • file (_io.TextIOWrapper) – Filepath to write the constraints to.

  • constraint (str) – Name of the restriction to add to the file.

  • terms (tuple) – Terms from the constraint equation.

  • target (bool, default False) – Features that are target variables get treated by write_descriptive_constraints function.

  • description (bool, default False) – Whether to include a description of the constraint to the file.

wip.modules.solver_ops.write_descriptive_constraints(file, model_target, datasets, df_detailed, scalers, models_coeficients, features_coeficient, models_results)[source]#

Write constraint built from the Ridge Regression model coefficients.

Model target → name of the selected model

Some target constraints have a different treatment when compared to other features.

Parameters

  • file (TextIOWrapper) – File to write the constraints

  • model_target (str) – Name of the constraint being written. This is the same name of the Ridge model.

  • datasets (Dict[str, pd.DataFrame]) – Dictionary with the datasets

  • df_detailed (pd.DataFrame) – Table with descriptions for each term of the constraint

  • scalers (Dict[str, sklearn.preprocessing.MinMaxScaler]) – Dictionary with the scalers for each tag (column)

  • models_coeficients (Dict[str, Dict[str, float]]) –

    Dictionary with the coefficients of each model. The dictionary should contain the following structure:

    {
    "model-name": {
        "tag": coefficient,
        # ...
    },
    "model-name-2": {
        "tag": coefficient,
        # ...
    },
    # ...
}

  • features_coeficient (zip) – List of tuples with the features and their coefficients

  • models_results (Dict[str, List[Dict[str, Any]]]) – Dictionary with the results of each model.

Returns

Dictionary with the coefficients of each model.

Return type

Dict[str, Dict[str, float]]

wip.modules.solver_ops.write_simple_constraints(file, model_target, models_results, unnormalize_constant)[source]#