def fit_transform(self, df, k=100):
'''
INPUT: pandas dataframe, int
OUTPUT: pandas dataframe
Fits internal variables used to fill missing values. Returns
transformed data ready for machine learning models.
'''
if self.method == 'mode':
self.fit_most_common_values(df)
elif self.method == 'kmeans':
self.max_age = max(df['Age'].values)
self.max_delta_h = max(df['Height (hh)']) - min(df['Height (hh)'])
print self.max_age, max(df['Height (hh)']), min(df['Height (hh)'])
kmeans = Kmeans_DF(df, k, self.horse_horse_distance)
self.centroids = kmeans.get_centroids()
print 'centroids fitted...'
return self.transform(df)
N = 8.
for col in ['Breed', 'Color', 'Pedigree', 'Sex']:
if (pd.isnull(series1[col])) or (pd.isnull(series2[col])):
N -= 1
elif series1[col] != series2[col]:
dist += 1
if not (pd.isnull(series1['Height (hh)']) or
pd.isnull(series2['Height (hh)'])):
dist += abs(series1['Height (hh)'] - series2['Height (hh)'])\
/ max_delta_h
else:
N -= 1
if not (pd.isnull(series1['Temperament']) or
pd.isnull(series2['Temperament'])):
dist += abs(series1['Temperament'] - series2['Temperament'])
else:
N -= 1
if not (pd.isnull(series1['Age']) or pd.isnull(series2['Age'])):
dist += abs(series1['Age'] - series2['Age']) / max_age
else:
N -= 1
return (dist / N - 1. / 8) * 8. / 7
if __name__ == '__main__':
for k in [50, 100, 200]:
filename = 'Centroids_' + str(k) + '.json'
kmeans = Kmeans_DF(df, k, horse_horse_distance)
centroids = kmeans.get_centroids()
centroids.to_json(filename)
print(str(k) + ' is done...')
