massive_interval_quality

hana_ml.algorithms.pal.stats.massive_interval_quality(data, significance_level, group_key=None, score_type=None, ave_abs_error=None, percent=None, check_consistency=None, thread_ratio=None)

Massive version of interval_quality().

This function calls SAP HANA PAL function PAL_MASSIVE_INTERVAL_QUALITY to evaluate interval forecast quality in parallel across multiple independent datasets identified by a group id.

Parameters
dataDataFrame

Input data for interval quality evaluation in massive mode.

This DataFrame must be structured as follows:

  • 1st column : GROUP ID, type INT, VARCHAR or NVARCHAR.

  • 2nd column : record ID, type INT, VARCHAR or NVARCHAR.

  • 3rd column : true value, type DOUBLE or DECIMAL(p,s).

  • 4th column : lower bound, type DOUBLE or DECIMAL(p,s).

  • 5th column : upper bound, type DOUBLE or DECIMAL(p,s).

significance_levelfloat

Significance level of prediction intervals. Must be in (0, 1).

group_keystr, optional

Name of the group id column. If not provided, it defaults to:

  • the first index column of data if data is indexed, otherwise

  • the first column of data.

score_type, ave_abs_error, percent, check_consistency, thread_ratio

Same meaning as in interval_quality().

Returns
DataFrames

DataFrame 1 : interval score result. Compared to single mode, it includes an additional group id column.

DataFrame 2 : statistics for interval score, with an additional group id column.

DataFrame 3 : error message table for massive execution.