Source code for surprise.accuracy

"""
The :mod:`surprise.accuracy` module provides with tools for computing accuracy
metrics on a set of predictions.

Available accuracy metrics:

.. autosummary::
    :nosignatures:

    rmse
    mae
    fcp
"""

from __future__ import (absolute_import, division, print_function,
                        unicode_literals)
from collections import defaultdict
import numpy as np

from .six import iteritems


[docs]def rmse(predictions, verbose=True): """Compute RMSE (Root Mean Squared Error). .. math:: \\text{RMSE} = \\sqrt{\\frac{1}{|\\hat{R}|} \\sum_{\\hat{r}_{ui} \in \\hat{R}}(r_{ui} - \\hat{r}_{ui})^2}. Args: predictions (:obj:`list` of :obj:`Prediction\ <surprise.prediction_algorithms.predictions.Prediction>`): A list of predictions, as returned by the :meth:`test <surprise.prediction_algorithms.bases.AlgoBase.test>` method. verbose: If True, will print computed value. Default is ``True``. Returns: The Root Mean Squared Error of predictions. Raises: ValueError: When ``predictions`` is empty. """ if not predictions: raise ValueError('Prediction list is empty.') mse = np.mean([float((true_r - est)**2) for (_, _, true_r, est, _) in predictions]) rmse_ = np.sqrt(mse) if verbose: print('RMSE: {0:1.4f}'.format(rmse_)) return rmse_
[docs]def mae(predictions, verbose=True): """Compute MAE (Mean Absolute Error). .. math:: \\text{MAE} = \\frac{1}{|\\hat{R}|} \\sum_{\\hat{r}_{ui} \in \\hat{R}}|r_{ui} - \\hat{r}_{ui}| Args: predictions (:obj:`list` of :obj:`Prediction\ <surprise.prediction_algorithms.predictions.Prediction>`): A list of predictions, as returned by the :meth:`test <surprise.prediction_algorithms.bases.AlgoBase.test>` method. verbose: If True, will print computed value. Default is ``True``. Returns: The Mean Absolute Error of predictions. Raises: ValueError: When ``predictions`` is empty. """ if not predictions: raise ValueError('Prediction list is empty.') mae_ = np.mean([float(abs(true_r - est)) for (_, _, true_r, est, _) in predictions]) if verbose: print('MAE: {0:1.4f}'.format(mae_)) return mae_
[docs]def fcp(predictions, verbose=True): """Compute FCP (Fraction of Concordant Pairs). Computed as described in paper `Collaborative Filtering on Ordinal User Feedback <http://www.ijcai.org/Proceedings/13/Papers/449.pdf>`_ by Koren and Sill, section 5.2. Args: predictions (:obj:`list` of :obj:`Prediction\ <surprise.prediction_algorithms.predictions.Prediction>`): A list of predictions, as returned by the :meth:`test <surprise.prediction_algorithms.bases.AlgoBase.test>` method. verbose: If True, will print computed value. Default is ``True``. Returns: The Fraction of Concordant Pairs. Raises: ValueError: When ``predictions`` is empty. """ if not predictions: raise ValueError('Prediction list is empty.') predictions_u = defaultdict(list) nc_u = defaultdict(int) nd_u = defaultdict(int) for u0, _, r0, est, _ in predictions: predictions_u[u0].append((r0, est)) for u0, preds in iteritems(predictions_u): for r0i, esti in preds: for r0j, estj in preds: if esti > estj and r0i > r0j: nc_u[u0] += 1 if esti >= estj and r0i < r0j: nd_u[u0] += 1 nc = np.mean(list(nc_u.values())) if nc_u else 0 nd = np.mean(list(nd_u.values())) if nd_u else 0 try: fcp = nc / (nc + nd) except ZeroDivisionError: raise ValueError('cannot compute fcp on this list of prediction. ' + 'Does every user have at least two predictions?') if verbose: print('FCP: {0:1.4f}'.format(fcp)) return fcp