def BordaCountMerge(self):
"""
return: a new DataFrame ranktable, with column=['title', 'rank']
"""
if self.cnt==0:
return pd.DataFrame(columns=['title', 'rate', 'rank'])
ranktable = self.ranktable
# ignore all the items that have nan as rank
ranktable = ranktable.dropna()
methods = ranktable.columns.drop('title')
candidate_score = np.zeros((ranktable.shape[0], ranktable.shape[1]-1),
np.int32)
for c in xrange(methods.shape[0]):
method = methods[c]
tmprank = ranktable[['title', method]].sort_values(by=method)
tmprank['_borda_score'] = pd.Series(
self._get_borda_score(tmprank.loc[:,method].values),
index=tmprank.index)
tmprank.sort_index(inplace=True)
candidate_score[:, c]=tmprank.loc[:,'_borda_score']
borda_score = candidate_score.sum(axis=1)
tmpitemlst = pd.DataFrame({'itemid': ranktable.loc[:,'title'],
'index': range(ranktable.shape[0])},
columns=['itemid', 'index'])
ranker = BaseRank(tmpitemlst)
rtn = ranker.rank(borda_score)
return rtn
def AverageRankMerge(self):
if self.cnt==0:
return pd.DataFrame(columns=['title', 'rate', 'rank'])
ranktable = self.ranktable
ranktable = ranktable.dropna()
methods = ranktable.columns.drop('title')
average_score = ranktable[methods].sum(axis=1).values
tmpitemlst = pd.DataFrame({'itemid': ranktable.loc[:,'title'],
'index': range(ranktable.shape[0])},
columns=['itemid', 'index'])
ranker = BaseRank(tmpitemlst, ascending=True)
rtn = ranker.rank(average_score)
return rtn
def same_rate_test():
ranker = BaseRank(sample_item_list, ascending=True)
ranktable = ranker.rank(sample_rate)
ranktable = ranktable[["title", "rate", "rank"]]
assert_array_equal(ranktable.iloc[:, 2].values, np.array([1, 1, 3], dtype=np.int32))
