df['pred'] = (df['oof'] >= 0.5) * 1
print(classification_report(df['gold'].values, df['pred'].values))
from sklearn.metrics import roc_auc_score
roc_auc_score(df['gold'].values, df['oof'].values)
thresholds = np.arange(0, 1, 0.001)
fscores = [
f1_score(df['gold'].values, (df['oof'] >= t) * 1) for t in thresholds
]
idx = np.argmax(fscores)
print(thresholds[idx], fscores[idx])
if __name__ == "__main__":
start = time.perf_counter()
make_print_to_file(path=config.SAVE_DIR)
# df_train = pd.read_csv(config.TRAIN_FILE, header=None, sep='\t')
# df_val = pd.read_csv(config.VAL_FILE, sep='\t')
# df_train.columns = ['tweet_id', 'user_id', 'tweet', 'label']
# run(df_train, df_val)
print(config.SEED)
run_fold(0)
run_fold(1)
run_fold(2)
run_fold(3)
run_fold(4)
# #
run_result()
test_prediction(kfold=config.KFOLD)
end = time.perf_counter()
df['pred'] = (df['oof'] >= 0.5) * 1
print(classification_report(df['gold'].values, df['pred'].values))
from sklearn.metrics import roc_auc_score
roc_auc_score(df['gold'].values, df['oof'].values)
thresholds = np.arange(0, 1, 0.001)
fscores = [
f1_score(df['gold'].values, (df['oof'] >= t) * 1) for t in thresholds
]
idx = np.argmax(fscores)
print(thresholds[idx], fscores[idx])
if __name__ == "__main__":
start = time.perf_counter()
make_print_to_file(path='.')
# df_train = pd.read_csv(config.TRAIN_FILE, header=None, sep='\t')
# df_val = pd.read_csv(config.VAL_FILE, sep='\t')
# df_train.columns = ['tweet_id', 'user_id', 'tweet', 'label']
# run(df_train, df_val)
print(config.SEED)
# run_fold(0)
# run_fold(1)
# run_fold(2)
# run_fold(3)
# run_fold(4)
# run_result()
test_prediction(kfold=config.KFOLD)
end = time.perf_counter()
time_cost = str((end - start) / 60)
print("time-cost:", time_cost)