#model=MLP(features_sizes,deep_layers=(256,256),k=256) #小batch=1024 LR不用小.同1e-3 valid_score=model.fit(train[features],valid[features],y_train,y_valid,lr=0.001,N_EPOCH=100,batch_size=1024,early_stopping_rounds=15)
#model = DeepFM(features_sizes, deep_layers=(256, 256), k=256)
#model = NFM(features_sizes, k=256)
#model = AFM(features_sizes,k=256,attention_FM=256)
#model = AFM(features_sizes, k=256, attention_FM=8,dropout_keeprate=0.9,lambda_l2=0.001)
#model = MLP(features_sizes, deep_layers=(1,), k=256)
#model=AutoInt(features_sizes,k=8)
valid_score = model.fit(train[features],
valid[features],
y_train,
y_valid,
lr=0.001,
N_EPOCH=100,
batch_size=4096,
early_stopping_rounds=15)
y_pred = model.predict(test[features]).reshape((-1))
predictions_bounded = np.maximum(y_pred,
np.ones(len(y_pred)) *
-1) # bound the lower values
predictions_bounded = np.minimum(predictions_bounded,
np.ones(len(y_pred)) *
1) # bound the higher values
test_loss = np.sqrt(
np.mean(
np.square(
y_test.reshape(predictions_bounded.shape) -
predictions_bounded)))
print("Protocol Test Score:", test_loss)
ls.append(test_loss)
if _ != Rounds - 1:
#model=CFM(features_sizes,loss_type='binary',metric_type='auc')
#model=MLR(features_sizes,loss_type='binary',metric_type='auc',MLR_m=16)
#model=MFM(features_sizes,k=8,loss_type='binary',metric_type='auc',MFM_m=2)
best_score = model.fit(X_train[cate_features],
X_valid[cate_features],
y_train,
y_valid,
lr=0.0005,
N_EPOCH=50,
batch_size=500,
early_stopping_rounds=3) #0.0005->0.001(1e-3 bs=1000)
SUBMIT = False
if SUBMIT:
y_pred = model.predict(test[cate_features])
y_pred = 1. / (1. + np.exp(-1. * y_pred))
sample_submission['isFraud'] = y_pred
#sample_submission.to_csv(data_path+'sub/sub01_LR_F49_timeSF_0.8154.csv',index=False)
#sample_submission.to_csv(data_path+'sub/sub05_MLR_m=15_nosig_F49_timeSF_0.8154.csv',index=False)
#LR:0.8774 KG:0.8261
#TIMESF
# LGB:0.8442@90 KG=0.8549
# CAT:with cat_fea
# dep=6:0.8471
# dep=8:0.8674 kg=0.8597
# LR:0.8154@11 KG=0.8468
# FM:0.8220@3 KG=0.8384 | FM k=6 0.8225 | FM k=4 0.8109 | k=3 0.8133
# MLP(256,128,64):0.8066@1 KG= | (64,64) 0.8022@1
# BiFM 0.8136 (very slow when fields a lot)