cybench.util package

Submodules

cybench.util.data module

cybench.util.data.data_to_pandas(data_items, data_cols=None)

Convert data items as dict to pandas DataFrame

Parameters:

  • data_items – list of data items, each of which is a dict

  • data_cols – list of keys to include as columns

Returns:

pd.DataFrame

cybench.util.data.trim_time_series_data(sample: dict, num_time_steps: int, time_series_keys: list)

Trims time series data to provided number of time steps

Parameters:

  • sample (dict) – key is name, value is np.ndarray

  • num_time_steps (int) – number of time steps to keep

  • time_series_keys (list) – keys for time series data

Returns:

the same sample with modified data

cybench.util.features module

cybench.util.features._add_period(df: DataFrame, period_length: str)

Add a period column.

Parameters:

  • df – pd.DataFrame

  • period_length – string, which can be “month”, “fortnight” or “dekad”

Returns:

pd.DataFrame

cybench.util.features._aggregate_by_period(df: DataFrame, index_cols: list, period_col: str, aggrs: dict, ft_cols: dict)

Aggregate data into features by period.

Parameters:

  • df – pd.DataFrame

  • index_cols – list of indices, which are location and year

  • period_col – string, column added by add_period()

  • aggrs – dict containing columns to aggregate (keys) and corresponding aggregation function (values)

  • ft_cols – dict for renaming columns to feature columns

Returns:

pd.DataFrame with features

cybench.util.features._count_threshold(df: DataFrame, index_cols: list, period_col: str, indicator: str, threshold_exceed: bool = True, threshold: float = 0.0, ft_name: str = None)

Aggregate data into features by period.

Parameters:

  • df – pd.DataFrame

  • index_cols – list of indices, which are location and year

  • period_col – string, column added by add_period()

  • indicator – string, indicator column to aggregate

  • threshold_exceed – boolean

  • threshold – float

  • ft_name – string name for aggregated indicator

Returns:

pd.DataFrame with features

cybench.util.features.dekad_from_date(dt: datetime)

Get the dekad number from date.

Parameters:

dt – date

Returns:

Dekad number, e.g. “YYYY0101” to “YYYY0110” -> 1,

”YYYY0111” to “YYYY0120” -> 2, “YYYY0121” to “YYYY0131” -> 3

cybench.util.features.design_features(crop: str, input_dfs: dict)

Design features based domain expertise.

Parameters:

  • crop (str) – crop name, e.g. maize

  • input_dfs (dict) – keys are input names, values are pd.DataFrames

Returns:

pd.DataFrame of features

cybench.util.features.fortnight_from_date(dt: datetime)

Get the fortnight number from date.

Parameters:

dt – date

Returns:

Fortnight number, “YYYY0101” to “YYYY0115” -> 1.

cybench.util.features.growing_degree_days(df: DataFrame, tbase: float)
cybench.util.features.unpack_time_series(df: DataFrame, indicators: list)

Unpack time series data to rows per date.

Parameters:

  • df – pd.DataFrame

  • indicators – list of indicators to unpack

Returns:

pd.DataFrame

cybench.util.torch module

cybench.util.torch.batch_tensors(*ts)

Module contents