File size: 5,216 Bytes
6077218 | 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 | import glob
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_statistical
from synthcity.plugins.core.dataloader import GenericDataLoader
pd.options.mode.chained_assignment = None
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--dataname', type=str)
parser.add_argument('--exp_name', type=str, default=None)
parser.add_argument('--non_learnable_schedule', action='store_true')
args = parser.parse_args()
def evaluate_quality(real_path, syn_path, info_path):
with open(info_path, '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')
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)
# Check for nan
if le_syn_data.isnull().values.any():
nan_coordinate = np.isnan(le_syn_data.to_numpy()).nonzero()
nan_row = np.unique(nan_coordinate[0])
print(f"Synthetic data contains NaN at row {nan_row}: ")
print(le_syn_data.iloc[nan_row])
return None, None
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']
return Alpha_Precision_all, Beta_Recall_all
if __name__ == '__main__':
exp_name = args.exp_name
if exp_name is None:
exp_name = "non_learnable_schedule" if args.non_learnable_schedule else "learnable_schedule"
dataname = args.dataname
data_dir = f'data/{dataname}'
info_path = f'{data_dir}/info.json'
real_path = f'synthetic/{dataname}/real.csv'
sample_dir = f"eval/report_runs/{exp_name}/{dataname}/all_samples"
sample_paths = glob.glob(os.path.join(sample_dir, "*.csv"))
print(f"{len(sample_paths )} samples loaded from {sample_dir}")
alphas, betas = [], []
for syn_path in sample_paths:
alpha_precision, beta_recall = evaluate_quality(real_path, syn_path, info_path)
if (alpha_precision is None) or (beta_recall is None):
continue
alphas.append(alpha_precision)
betas.append(beta_recall)
alphas = np.array(alphas)
betas = np.array(betas)
alpha_percent = alphas * 100
beta_percent = betas * 100
quality = pd.DataFrame({
'alpha': alpha_percent,
'beta': beta_percent
})
avg = quality.mean(axis=0).round(2)
std = quality.std(axis=0).round(2)
quality_avg_std = pd.concat([avg, std], axis=1, ignore_index=True)
quality_avg_std.columns = ["avg", "std"]
quality_avg_std.index = ["alpha", "beta"]
save_dir = os.path.dirname(sample_dir)
quality.to_csv(os.path.join(save_dir, "quality.csv"), index=True)
avg_std = pd.read_csv(os.path.join(save_dir, "avg_std.csv"), index_col=0)
avg_std = pd.concat([avg_std, quality_avg_std])
print(avg_std)
avg_std.to_csv(os.path.join(save_dir, "avg_std.csv"), index=True)
|