Spaces:
Running
Running
File size: 8,130 Bytes
c4ac745 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 | import numpy as np
import pandas as pd
import os
import sys
import json
sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from sklearn.preprocessing import OneHotEncoder
from synthcity.metrics import eval_detection, eval_performance, eval_statistical
from synthcity.plugins.core.dataloader import GenericDataLoader
pd.options.mode.chained_assignment = None
import argparse
def eval_metrics(syn_data, real_data, info):
real_data.columns = range(len(real_data.columns))
syn_data.columns = range(len(syn_data.columns))
num_col_idx = info['num_col_idx']
cat_col_idx = info['cat_col_idx']
target_col_idx = info['target_col_idx']
if info['task_type'] == 'regression':
num_col_idx += target_col_idx
else:
cat_col_idx += target_col_idx
num_real_data = real_data[num_col_idx]
cat_real_data = real_data[cat_col_idx]
num_real_data_np = num_real_data.to_numpy()
cat_real_data_np = cat_real_data.to_numpy().astype('str')
num_syn_data = syn_data[num_col_idx]
cat_syn_data = syn_data[cat_col_idx]
num_syn_data_np = num_syn_data.to_numpy()
# cat_syn_data_np = np.array
if cat_real_data.shape[1] > 0:
cat_syn_data_np = cat_syn_data.to_numpy().astype('str')
encoder = OneHotEncoder()
encoder.fit(np.concatenate((cat_real_data_np, cat_syn_data_np), axis=0))
cat_real_data_oh = encoder.transform(cat_real_data_np).toarray()
cat_syn_data_oh = encoder.transform(cat_syn_data_np).toarray()
else:
cat_real_data_oh = np.empty((cat_real_data.shape[0], 0))
cat_syn_data_oh = np.empty((cat_syn_data.shape[0], 0))
le_real_data = pd.DataFrame(np.concatenate((num_real_data_np, cat_real_data_oh), axis = 1)).astype(float)
le_syn_data = pd.DataFrame(np.concatenate((num_syn_data_np, cat_syn_data_oh), axis = 1)).astype(float)
np.set_printoptions(precision=4)
print('=========== All Features ===========')
print('Data shape: ', le_syn_data.shape)
X_syn_loader = GenericDataLoader(le_syn_data)
X_real_loader = GenericDataLoader(le_real_data)
quality_evaluator = eval_statistical.AlphaPrecision()
qual_res = quality_evaluator.evaluate(X_real_loader, X_syn_loader)
qual_res = {
k: v for (k, v) in qual_res.items() if "naive" in k
} # use the naive implementation of AlphaPrecision
# qual_score = np.mean(list(qual_res.values()))
print('alpha precision: {:.6f}, beta recall: {:.6f}'.format(qual_res['delta_precision_alpha_naive'], qual_res['delta_coverage_beta_naive'] ))
Alpha_Precision_all = qual_res['delta_precision_alpha_naive']
Beta_Recall_all = qual_res['delta_coverage_beta_naive']
# save_dir = f'eval/quality/{dataname}'
# if not os.path.exists(save_dir):
# os.makedirs(save_dir)
# with open(f'{save_dir}/{model}.txt', 'w') as f:
# f.write(f'{Alpha_Precision_all}\n')
# f.write(f'{Beta_Recall_all}\n')
return Alpha_Precision_all, Beta_Recall_all
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--dataname', type=str, default='adult')
parser.add_argument('--model', type=str, default='model')
parser.add_argument('--path', type=str, default = None, help='The file path of the synthetic data')
args = parser.parse_args()
dataname = args.dataname
model = args.model
if not args.path:
syn_path = f'synthetic/{dataname}/{model}.csv'
else:
syn_path = args.path
real_path = f'synthetic/{dataname}/real.csv'
data_dir = f'data/{dataname}'
print(syn_path)
with open(f'{data_dir}/info.json', 'r') as f:
info = json.load(f)
syn_data = pd.read_csv(syn_path)
real_data = pd.read_csv(real_path)
''' Special treatment for default dataset and CoDi model '''
real_data.columns = range(len(real_data.columns))
syn_data.columns = range(len(syn_data.columns))
num_col_idx = info['num_col_idx']
cat_col_idx = info['cat_col_idx']
target_col_idx = info['target_col_idx']
if info['task_type'] == 'regression':
num_col_idx += target_col_idx
else:
cat_col_idx += target_col_idx
num_real_data = real_data[num_col_idx]
cat_real_data = real_data[cat_col_idx]
num_real_data_np = num_real_data.to_numpy()
cat_real_data_np = cat_real_data.to_numpy().astype('str')
num_syn_data = syn_data[num_col_idx]
cat_syn_data = syn_data[cat_col_idx]
num_syn_data_np = num_syn_data.to_numpy()
# cat_syn_data_np = np.array
cat_syn_data_np = cat_syn_data.to_numpy().astype('str')
if (dataname == 'default' or dataname == 'news') and model[:4] == 'codi':
cat_syn_data_np = cat_syn_data.astype('int').to_numpy().astype('str')
elif model[:5] == 'great':
if dataname == 'shoppers':
cat_syn_data_np[:, 1] = cat_syn_data[11].astype('int').to_numpy().astype('str')
cat_syn_data_np[:, 2] = cat_syn_data[12].astype('int').to_numpy().astype('str')
cat_syn_data_np[:, 3] = cat_syn_data[13].astype('int').to_numpy().astype('str')
max_data = cat_real_data[14].max()
cat_syn_data.loc[cat_syn_data[14] > max_data, 14] = max_data
# cat_syn_data[14] = cat_syn_data[14].apply(lambda x: threshold if x > max_data else x)
cat_syn_data_np[:, 4] = cat_syn_data[14].astype('int').to_numpy().astype('str')
cat_syn_data_np[:, 4] = cat_syn_data[14].astype('int').to_numpy().astype('str')
elif dataname in ['default', 'faults', 'beijing']:
columns = cat_real_data.columns
for i, col in enumerate(columns):
if (cat_real_data[col].dtype == 'int'):
max_data = cat_real_data[col].max()
min_data = cat_real_data[col].min()
cat_syn_data.loc[cat_syn_data[col] > max_data, col] = max_data
cat_syn_data.loc[cat_syn_data[col] < min_data, col] = min_data
cat_syn_data_np[:, i] = cat_syn_data[col].astype('int').to_numpy().astype('str')
else:
cat_syn_data_np = cat_syn_data.to_numpy().astype('str')
else:
cat_syn_data_np = cat_syn_data.to_numpy().astype('str')
encoder = OneHotEncoder()
encoder.fit(cat_real_data_np)
cat_real_data_oh = encoder.transform(cat_real_data_np).toarray()
cat_syn_data_oh = encoder.transform(cat_syn_data_np).toarray()
le_real_data = pd.DataFrame(np.concatenate((num_real_data_np, cat_real_data_oh), axis = 1)).astype(float)
le_real_num = pd.DataFrame(num_real_data_np).astype(float)
le_real_cat = pd.DataFrame(cat_real_data_oh).astype(float)
le_syn_data = pd.DataFrame(np.concatenate((num_syn_data_np, cat_syn_data_oh), axis = 1)).astype(float)
le_syn_num = pd.DataFrame(num_syn_data_np).astype(float)
le_syn_cat = pd.DataFrame(cat_syn_data_oh).astype(float)
np.set_printoptions(precision=4)
result = []
print('=========== All Features ===========')
print('Data shape: ', le_syn_data.shape)
X_syn_loader = GenericDataLoader(le_syn_data)
X_real_loader = GenericDataLoader(le_real_data)
quality_evaluator = eval_statistical.AlphaPrecision()
qual_res = quality_evaluator.evaluate(X_real_loader, X_syn_loader)
qual_res = {
k: v for (k, v) in qual_res.items() if "naive" in k
} # use the naive implementation of AlphaPrecision
qual_score = np.mean(list(qual_res.values()))
print('alpha precision: {:.6f}, beta recall: {:.6f}'.format(qual_res['delta_precision_alpha_naive'], qual_res['delta_coverage_beta_naive'] ))
Alpha_Precision_all = qual_res['delta_precision_alpha_naive']
Beta_Recall_all = qual_res['delta_coverage_beta_naive']
save_dir = f'eval/quality/{dataname}'
if not os.path.exists(save_dir):
os.makedirs(save_dir)
with open(f'{save_dir}/{model}.txt', 'w') as f:
f.write(f'{Alpha_Precision_all}\n')
f.write(f'{Beta_Recall_all}\n')
|