Group pairwise comparison

Group pairwise object and calculations. See pyanp.priority for all methods of calculating priorities from a pairwise comparison matrix in addition to inconsistency calculations.

class pyanp.pairwise.Pairwise(alts=None, users=None, demographic_cols=None)[source]

Creates a new group pairwise comparison object.

Parameters:
  • alts – The list alternatives (things you are comparing) to start with. Should be a list-like object of strings.
  • users – The users to start the group pairwise comparison object with. It should be a list-like object of strings.
  • demographic_cols – The names of the demographic columns to start the group pairwise comparison object with. It should be a list-like object of strings.
add_alt(alt_name: str, ignore_existing=False) → None[source]

Adds an alternative (thing you are pairwise comparing) to this group pairwise comparison object.

Parameters:alt_name – The name of the alternative to add.
Returns:Nothing
Raises:ValueError – If the alternative already esisted.
add_user(user_name: str) → None[source]

Adds a user to this group pairwise comparison object.

Parameters:user_name – The name of the user to add
Returns:Nothing
Raises:ValueError – If the user already existed.
alt_index(alt_name_or_index) → int[source]

Find the index (integer location) of the given alternative in the pairwise comparison matrices.

Parameters:alt_name_or_index – If this is an integer, we simply return that integer. Otherwise we look up the index of the alternative name in the list of alternatives in this object.
Returns:The index that alternative has in the pairwise comparison matrices.
alt_names()[source]
Returns:List of string alt names
data_names(append_to=None, post_pend='')[source]
incon_std(user_name=None) → float[source]

Calculates the standard Saaty pairwise comparison inconsistency for a user or group of users.

Parameters:user_name – The name/names of the users to get the inconsistency of. If None, we get the inconsistency of the group average matrix. If it is a string, we get the inconsistency of that user. If it is a list of users, we get the inconsistency of the group average for that list of users.
Returns:The Saaty inconsistency score.
is_alt(alt_name: str) → bool[source]

Checks if an alternative (a thing you are pairwise comparing) exists in this group pairwise comparison object.

Parameters:alt_name – The name of the alternative to check for.
Returns:True/False
is_user(user_name: str) → bool[source]

Checks if a user exists in this group pairwise comparison object.

Parameters:user_name – The name of the user to look for
Returns:True/False
matrix(user_name=None, createUnknownUser: bool = True, as_df=False) → numpy.ndarray[source]

Gets the pairwise comparison for a user or group of users.

Parameters:
  • user_name – The name/names of the user/users to get the comparisons of. If None, that means to get the group average for all users. If it is a string, that means get the pairwise comparison matrix of that user. If it is a list-like of strings, we get the group average matrix for all of those users.
  • createUnknownUser – If True and the user_name did not exist, we should create that user. Otherwise throw an error if we request for a non-existant user.
  • as_df – If True return as pandas.DataFrame with index/column names as the alt names, otherwise return numpy.ndarray
Returns:

The numpy array of the pairwise comparisons, or DataFrame if as_df is True
Raises:

ValueError – If createUnknownUser=False and we request for a single non-existant user.
nalts() → int[source]
Returns:The number of alternatives (things you are pairwise comparing) in this group pairwise comparison object.
priority(username=None, ptype: pyanp.prioritizer.PriorityType = None)[source]

Calculates the resulting priority for the given user / users.

Parameters:
  • user_name – The name/names of the users to calculate the priority of. If None, we get the priority of the group average matrix. If it is a string, we get the priority of that user. If it is a list of users, we get the priority of the group average for that list of users.
  • ptype – How should we normalize the resulting priorities (if at all).
Returns:

A pandas.Series whose indices are the alternative names and whose values are the priorities of those alternatives.
unvote(user_name: str, row, col, createUnknownUser: bool = True) → None[source]

Unsets a pairwise comparison

Parameters:
  • user_name – The string name of the user whose pairwise comparison vote you wish to unset.
  • row – The integer or string name of the row to compare at.
  • col – The integer or string name of the column to compare at.
  • createUnknownUser – If True and user_name does not exist in this object, we will create it first, then do the unset operation. Otherwise it throws an exception for unknown users.
Returns:

Nothing
Raises:

ValueError – If the user does not exist and createUnknownUsers is False.
usernames()[source]
Returns:A list of the users in this group pairwise comparison object.
vote(user_name: str, row, col, val: float = 0, createUnknownUser: bool = True) → None[source]

Changes a single pairwise value for a single user.

Parameters:
  • user_name – The string name of the user whose pairwise comparison vote you wish to change.
  • row – The integer or string name of the row to compare at.
  • col – The integer or string name of the column to compare at.
  • val – The new pairwise comparison value
  • createUnknownUser – If True and user_name does not exist in this object, we will create it first, then do the comparison. Otherwise it throws an exception for unknown users.
Returns:

Nothing
Raises:

ValueError – If the user does not exist and createUnknownUsers is False.
vote_matrix(user_name: str, val=<class 'numpy.ndarray'>, createUnknownUser: bool = True)[source]

Sets the vote matrix for a user

Parameters:
  • user_name –
  • val –
Returns:
vote_series(votes: pandas.core.series.Series, row, col, createUnknownUser: bool = True) → None[source]

Changes a single pairwise value for a series of users.

Parameters:
  • votes – Series whose index is usernames and values are their votes.
  • row – The integer or string name of the row to compare at.
  • col – The integer or string name of the column to compare at.
  • createUnknownUser – If True and a username does not exist in this object, we will create it first, then do the comparison. Otherwise it throws an exception for unknown users.
Returns:

Nothing
Raises:

ValueError – If the user does not exist and createUnknownUsers is False.
pyanp.pairwise.add_place(mat)[source]

Adds a row and column to the end of a matrix, and makes the last entry 1, rest of the added entries are zeroes

Parameters:mat – The matrix to add an entry to.
Returns:New matrix
pyanp.pairwise.geom_avg_mats(mats) → numpy.ndarray[source]

Calculates the geometric average of the given matrices.

Parameters:mats – A list-like object of numpy arrays
Returns:A numpy array that is the geometric average